CROSS-REFERENCE TO RELATED APPLICATION

[0001] This application claims the benefit of priority of Singapore Patent Application No. 10202250468P, filed on 14 July 2022, the content of which being hereby incorporated by reference in its entirety for all purposes.

TECHNICAL FIELD

[0002] The present invention generally relates to monostatic sensing over Wi-Fi communication, including a sensing system and a method for monostatic sensing over Wi-Fi communication.

BACKGROUND

[0003] Although the Received Signal Strength (RSS) carried by Wi-Fi signaling has been exploited by indoor localization for more than two decades, the true Wi-Fi sensing (i.e., leveraging Wi-Fi communications) only started about a decade ago thanks to the ability of extracting Channel State Information (CSI) from data packets. In particular, numerous applications of device-free Wi-Fi sensing (i.e., Wi-Fi sensing without requiring the subject/object to be sensed to have an associated Wi-Fi device) have been proposed to utilize CSI, notably including indoor localization, activity/gesture recognition, vital signs monitoring, and object identification/imaging. Although these applications all bear a promising future, the bistatic or multistatic nature of Wi-Fi infrastructure has largely hampered the deployment of real-life or practical systems. Basically, as a Wi-Fi communication session involves at least a pair of transmitter (Tx) and receiver (Rx), any sensing function piggybacking on such an infrastructure is conventionally subject to the severe constraints imposed by the physical separation of Tx and Rx.

[0004] Among all constraints imposed by Wi-Fi’s bistatic or multistatic nature, three prominent constraints are illustrated in FIG. 1 by way of examples only and without limitations. In particular, FIG. 1 depicts a schematic drawing illustrating conventional Wi-Fi sensing, whereby the thin arrows represent various RF propagations originating from the transmitter (Tx) along various distinct paths, while the thick (double-sided) arrows denote the directions of sensed subject motions. [0005] First of all, given the uncertainties, such as the existence of carrier frequency offset and the lack of synchronization between Tx and Rx, even estimating the time-of-flight (ToF) of the line-of-sight (LoS) path between Tx and Rx entails a very cumbersome process. Unfortunately, the ToFs of the non-LoS (NLoS) paths, albeit essential to device-free sensing, simply cannot be estimated. Moreover, although estimating angle-of-arrival (AoA) becomes the centre of Wi-Fi sensing due to the inability of obtaining ToF, the strong signal of the (useless) LoS path could easily overwhelm the essential NLoS paths. Consequently, existing Wi-Fi sensing techniques often have to rely on multiple Wi-Fi links (i.e., multistatic setting) and/or on motion effect as an extra hint. Finally, it is well known that the motion effect captured by a reflected RF signal represents the distance variation of a reflecting subject along a certain direction. As illustrated in FIG. 1, this direction happens to be the gradient of the Fresnel field, yet this gradient (along which the reflection path length changes) varies with the (unknown) location of the reflecting subject due to the bistatic nature of Wi-Fi, thus causing ambiguity in interpreting the motion sensing results.

[0006] A need therefore exists to provide monostatic sensing over Wi-Fi communication, including a sensing system and a method for monostatic sensing over Wi-Fi communication, that seek to overcome, or at least ameliorate, one or more deficiencies or constraints in bistatic or multistatic sensing over Wi-Fi communication, and more particularly, in an effective and efficient manner. It is against this background that the present invention has been developed.

SUMMARY

[0007] According to a first aspect of the present invention, there is provided a sensing system for monostatic sensing over Wi-Fi communication, the sensing system comprising: an analog cancellator configured to receive an input analog signal comprising a transmission signal component and a plurality of multi-path reflection signal components and to suppress the transmission signal component of the input analog signal for producing a transmission signal suppressed analog signal comprising a residue transmission signal component; a digital cancellator configured to receive the transmission signal suppressed analog signal and to suppress the residue transmission signal component for producing a transmission signal suppressed digital signal; a radio frequency switch configured to selectively switch a coupling to a communication port from between an antenna and a predetermined load for facilitating optimization of a parameter associated with the analog cancellator for suppressing the transmission signal component of the input analog signal and a parameter associated with the digital cancellator for suppressing the residue transmission signal component; and a channel state information generator configured to generate channel state information for monostatic sensing based on the transmission signal suppressed digital signal.

[0008] According to a second aspect of the present invention, there is provided a method of monostatic sensing over Wi-Fi communication, the method comprising: receiving, by an analog cancellator, an input analog signal comprising a transmission signal component and a plurality of multi-path reflection signal components; suppressing, by the analog cancellator, the transmission signal component of the input analog signal to produce a transmission signal suppressed analog signal comprising a residue transmission signal component; receiving, by a digital cancellator, the transmission signal suppressed analog signal; suppressing, by the digital cancellator, the residue transmission signal component to produce a transmission signal suppressed digital signal; and generating, by a channel state information generator, channel state information for monostatic sensing based on the transmission signal suppressed digital signal, wherein the method further comprises selectively switching, using a radio frequency switch, a coupling to a communication port from between an antenna and a predetermined load for facilitating optimization of a parameter associated with the analog cancellator for suppressing the transmission signal component of the input analog signal and a parameter associated with the digital cancellator for suppressing the residue transmission signal component.

[0009] According to a third aspect of the present invention, there is provided a method of forming a sensing system for monostatic sensing over Wi-Fi communication, the method comprising: providing an analog cancellator configured to receive an input analog signal comprising a transmission signal component and a plurality of multi-path reflection signal components and to suppress the transmission signal component of the input analog signal for producing a transmission signal suppressed analog signal comprising a residue transmission signal component; providing a digital cancellator configured to receive the transmission signal suppressed analog signal and to suppress the residue transmission signal component for producing a transmission signal suppressed digital signal; providing a radio frequency switch configured to selectively switch a coupling to a communication port from between an antenna and a predetermined load for facilitating optimization of a parameter associated with the analog cancellator for suppressing the transmission signal component of the input analog signal and a parameter associated with the digital cancellator for suppressing the residue transmission signal component; and providing a channel state information generator configured to generate channel state information for monostatic sensing based on the transmission signal suppressed digital signal.

BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Embodiments of the present invention will be better understood and readily apparent to one of ordinary skill in the art from the following written description, by way of example only, and in conjunction with the drawings, in which:

FIG. 1 depicts a schematic drawing illustrating conventional Wi-Fi sensing, whereby the thin arrows represent various RF propagations originating from the transmitter (Tx) along various distinct paths, while the thick (double-sided) arrows denote the directions of sensed subject motions;

FIG. 2 depicts a schematic drawing of a sensing system for monostatic sensing over WiFi communication, according to various embodiments of the present invention;

FIG. 3 depicts a schematic flow diagram of a method of monostatic sensing over Wi-Fi communication, according to various embodiments of the present invention;

FIG. 4 depicts a schematic flow diagram of a method of forming a sensing system for monostatic sensing over Wi-Fi communication, according to various embodiments of the present invention;

FIG. 5 depicts a schematic drawing illustrating ISAC-Fi (integrated sensing and communication over Wi-Fi) sensing according to various example embodiments of the present invention, whereby the thin arrows represent various RF propagations originating from different transmitters (Tx) along various distinct paths, while the thick (double-sided) arrows denote the directions of sensed subject motions;

FIG. 6A depicts a schematic drawing of an example architecture of a conventional WiFi network interface card (NIC); FIG. 6B depicts a schematic drawing of an example architecture of a full version of ISAC-Fi, according to various example embodiments of the present invention;

FIG. 6C depicts a schematic drawing of an example architecture of the partial ISAC-Fi, according to various example embodiments of the present invention;

FIGs. 7A and 7B show plots of the unwrapped CSI phases of 52 subcarriers and across consecutive symbols under bistatic mode and monostatic mode, respectively;

FIGs. 8A and 8B depict phase variations induced by human (target) breath at two different Rx-target ranges under bistatic mode and monostatic mode, respectively;

FIG. 9 shows the conceptual illustration and experiment settings for the sensing motion effect under bistatic mode;

FIGs. 10A and 10B show plots of the unwrapped CSI phases of the 1st subcarrier when a motor-driving slide rail is placed perpendicular (FIG. 10A) and parallel (FIG. 10B) to the Tx- Rx line;

FIG. 11A depicts a schematic drawing of an example first or full version of ISAC-Fi with Tx-Rx signal separation, according to various example embodiments of the present invention;

FIG. 1 IB depicts a schematic drawing of an example second or partial version of ISAC- Fi with Tx-Rx signal separation, according to various example embodiments of the present invention;

FIG. 12 depicts a schematic drawing of an analog cancellator implemented in the form of a Direct Quadrature Modulator (DQM) and communicatively coupled to a circulator (in the case of the full ISAC-Fi) or a hybrid coupler (in the case of the partial ISAC-Fi), according to various example embodiments of the present invention;

FIG. 13 illustrates the erasing of the monostatic sensing signal by analog and digital cancellators without self-adapted Tx-Rx signal separation;

FIG. 14A shows a plot of the correlation coefficients of Tx-interferences under two antenna settings, according to various example embodiments of the present invention;

FIG. 14B shows a plot of the correlation coefficients of a Rx signal right after calibrations and those received later, according to various example embodiments of the present invention;

FIG. 15 shows a plot of the human breath signals with and without SA (self-adapted Tx- Rx signal separation), according to various example embodiments of the present invention; FIGs. 16A and 16B show plots illustrating the self-adapted Tx-Rx signal separator (SA) heavily degrading normal Wi-Fi packet reception quality in terms of both SNR (FIG. 16A) and throughput (FIG. 16B);

FIGs. 17A and 17B show the STFT (short-time Fourier transform) heatmaps of a human slowly walking under regular (FIG. 17 A) and irregular packets (FIG. 17B);

FIG. 18 shows plots of the power spectrum of the received baseband signal after various components of Tx-Rx signal separators, according to various example embodiments of the present invention;

FIGs. 19A and 19B show the impact of the Tx-Rx signal separation on normal Wi-Fi communication, according to various example embodiments of the present invention;

FIGs. 20A and 20B show the ranging errors obtained for the full and partial ISAC-Fi versions, respectively, for comparison, according to various example embodiments of the present invention;

FIGs. 21 A and 21B show the the velocity errors of ISAC-Fi and FFT for the full and partial ISAC-Fi versions, respectively, according to various example embodiments of the present invention;

FIGs. 22 A and 22B show the performance of non-collaborative (ISAC-Fi i) and collaborative (ISAC-Fi2) MIMO localization, according to various example embodiments of the present invention;

FIGs. 23 A and 23B show the confusion matrices of human activity recognition (HAR) of ISAC-Fi according to various example embodiments and the environment independent (El) framework, according to various example embodiments of the present invention; and

FIG. 24 shows the imaging results of human subjects obtained by the ISAC-Fi, according to various example embodiments of the present invention, including the ground truth photos, RF outlines, and RF skeletons of one, two, and three subjects, respectively.

DETAILED DESCRIPTION

[0011] Various embodiments of the present invention relate to monostatic sensing over WiFi communication, including a sensing system and a method for monostatic sensing over WiFi communication.

[0012] As discussed in the background, although Wi-Fi communications have previously been exploited for sensing purposes, the bistatic or multistatic nature of Wi-Fi still poses multiple challenges, hampering the deployment of real-life or practical systems. Basically, as a Wi-Fi communication session involves at least a pair of transmitter (Tx) and receiver (Rx), any sensing function piggybacking on such an infrastructure is conventionally subject to the severe constraints imposed by the physical separation of Tx and Rx, such as those constraints illustrated and discussed in the background with reference to FIG. 1. Accordingly, various embodiments of the present invention provide monostatic sensing over Wi-Fi communication, including a sensing system and a method for monostatic sensing over Wi-Fi communication, that seek to overcome, or at least ameliorate, one or more deficiencies or constraints in bistatic or multistatic sensing over Wi-Fi communication, and more particularly, in an effective and efficient manner.

[0013] FIG. 2 depicts a schematic drawing of a sensing system 200 for monostatic sensing over Wi-Fi communication, according to various embodiments of the present invention. The sensing system 200 comprising: an analog cancellator 206 configured to receive an input analog signal comprising a transmission signal component and a plurality of multi-path reflection signal components and to suppress the transmission signal component of the input analog signal for producing a transmission signal suppressed analog signal comprising a residue transmission signal component; a digital cancellator 208 configured to receive the transmission signal suppressed analog signal and to suppress the residue transmission signal component for producing a transmission signal suppressed digital signal; a radio frequency (RF) switch 210 configured to selectively switch a coupling to a communication port from between an antenna 212 and a predetermined load 214 (i.e., selectively switching a coupling between connecting the antenna 212 to the communication port and connecting the predetermined load 214 to the communication port) for facilitating optimization of a parameter associated with the analog cancellator 206 for suppressing the transmission signal component of the input analog signal and a parameter associated with the digital cancellator 208 for suppressing the residue transmission signal component; and a channel state information generator 216 configured to generate channel state information for monostatic sensing based on the transmission signal suppressed digital signal.

[0014] Accordingly, the sensing system 200 according to various embodiments of the present invention is advantageously able to perform monostatic sensing over Wi-Fi communication in an effective and efficient manner. In particular, not only does the sensing system 200 comprise an analog cancellator 206 and a digital cancellator 208 configured to, in series/stages, remove undesired transmission signal interference in a signal received (comprising a combination of transmission signal as undesired interference and multi-path reflection signals as desired sensing signals for monostatic sensing), the sensing system 200 further comprises a RF switch 210 for facilitating optimization of parameters associated with the analog cancellator 206 and the digital cancellator 208 for improving or optimizing the removal of the undesired transmission signal interference in the signal received. Therefore, the sensing system 200 according to various embodiments of the present invention is advantageously able to perform monostatic sensing over Wi-Fi communication in an effective and efficient manner. These advantages or technical effects, and/or other advantages or technical effects, will become more apparent to a person skilled in the art as the sensing system 200 for monostatic sensing over Wi-Fi communication, as well as the corresponding method of monostatic sensing over Wi-Fi communication, is described in more detail according to various embodiments and example embodiments of the present invention.

[0015] In various embodiments, the parameter associated with the analog cancellator 206 is a channel gain of the analog cancellator 206. In this regard, the parameter associated with the analog cancellator 206 is optimized with respect to a channel gain of a hardware of the sensing system 200 for suppressing the transmission signal component of the input analog signal.

[0016] In various embodiments, the analog cancellator 206 comprises a direct quadrature modulator, and an IQ baseband generator of the direct quadrature modulator is configured to adjust the parameter associated with the analog cancellator 206 to match the channel gain of the hardware of the sensing system 200.

[0017] In various embodiments, the digital cancellator 208 comprises an adaptive filter, and coefficients of the adaptive filter are configured to suppress the residue transmission signal component. In this regard, the parameter associated with the digital cancellator 208 corresponds to the coefficients of the adaptive filter.

[0018] In various embodiments, the adaptive filter is configured to produce a linear combination of multiple time-delayed versions of a baseband of a reference preamble (e.g., a pre-stored and known Wi-Fi preamble) based on the coefficients of the adaptive filter for suppressing the residue transmission signal component.

[0019] In various embodiments, the channel state information generator 216 is configured to: estimate channel features of a plurality of transmission signal suppressed digital signals produced respectively for a plurality of input analog signals received, including performing sparse optimization based on the plurality of transmission signal suppressed digital signals including irregular data packets. [0020] In various first embodiments, the sensing system 200 comprises a network interface card comprising the analog cancellator 206, the digital cancellator 208 and the channel state information generator 216. In this regard, the analog cancellator 206 and the digital cancellator 208 are each configured to receive a control signal for respectively controlling an operation thereof.

[0021] In various first embodiments, the network interface card is configured to be transitionable amongst a plurality of states based on a type of a message received or sent. In this regard, the control signal for the analog cancellator 206 and the digital cancellator 208 corresponds to the message received or sent.

[0022] In various first embodiments, the plurality of states comprise a communication state, a monostatic sensing state and a bistatic sensing state. The network interface card is configured to be in the monostatic sensing state, and the analog cancellator 206 and the digital cancellator 208 are controlled to be enabled based on the message indicating data transmission or acknowledgement of a predetermined type of data reception.

[0023] In various second embodiments, the sensing system 200 comprises a sensing device comprising the analog cancellator 206, the digital cancellator 208 and the channel state information generator 216. In this regard, the sensing device is removably connectable to a WiFi network interface card.

[0024] In various second embodiments, the sensing device is a software-defined radio (SDR).

[0025] FIG. 3 depicts a schematic flow diagram of a method 300 of monostatic sensing over Wi-Fi communication, according to various embodiments of the present invention. The method 300 comprises: receiving (at 302), by an analog cancellator 206, an input analog signal comprising a transmission signal component and a plurality of multi-path reflection signal components; suppressing (at 304), by the analog cancellator 206, the transmission signal component of the input analog signal to produce a transmission signal suppressed analog signal comprising a residue transmission signal component; receiving (at 306), by a digital cancellator 208, the transmission signal suppressed analog signal; suppressing (at 308), by the digital cancellator 208, the residue transmission signal component to produce a transmission signal suppressed digital signal; and generating (at 310), by a channel state information generator 216, channel state information for monostatic sensing based on the transmission signal suppressed digital signal. In particular, the method 300 further comprises selectively switching (at 312), using a radio frequency switch 210, a coupling to a communication port from between an antenna 212 and a predetermined load 214 for facilitating optimization of a parameter associated with the analog cancellator 206 for suppressing the transmission signal component of the input analog signal and a parameter associated with the digital cancellator 208 for suppressing the residue transmission signal component.

[0026] In various embodiments, the method 300 of monostatic sensing corresponds to the sensing system 200 as described hereinbefore according to various embodiments with reference to FIG. 2, therefore, various steps or operations of the method 300 may correspond to various functions or operations in which various components (or elements) of the sensing system 200 are configured to perform, as described hereinbefore according to various embodiments. In other words, various embodiments described herein in context of the sensing system 200 are analogously valid for the method 300 of monostatic sensing, and vice versa.

[0027] In various embodiments, the parameter associated with the analog cancellator 206 is a channel gain of the analog cancellator 206. In this regard, the parameter associated with the analog cancellator 206 is optimized with respect to a channel gain of a hardware of the sensing system for suppressing the transmission signal component of the input analog signal.

[0028] In various embodiments, the analog cancellator 206 comprises a direct quadrature modulator. In this regard, the above-mentioned suppressing (at 304), by the analog cancellator 206, the transmission signal component of the input analog signal comprises adjusting, by an IQ baseband generator of the direct quadrature modulator, the parameter associated with the analog cancellator 206 to match the channel gain of the hardware of the sensing system 200 for suppressing the transmission signal component.

[0029] In various embodiments, the digital cancellator 208 comprises an adaptive filter, and coefficients of the adaptive filter are configured to suppress the residue transmission signal component. In this regard, the parameter associated with the digital cancellator 208 corresponds to the coefficients of the adaptive filter.

[0030] In various embodiments, the above-mentioned suppressing (at 308), by the digital cancellator 208, the residue transmission signal component comprises producing, by the adaptive filter, a linear combination of multiple time-delayed versions of a baseband of a reference preamble based on the coefficients of the adaptive filter for suppressing the residue transmission signal component.

[0031] In various embodiments, the above-mentioned generating (at 310), by the channel state information generator 216, channel state information comprises estimating channel features of a plurality of transmission signal suppressed digital signals produced respectively for a plurality of input analog signals received, including performing sparse optimization based on the plurality of transmission signal suppressed digital signals including irregular data packets.

[0032] In various first embodiments, the analog cancellator 206, the digital cancellator 208 and the channel state information generator 216 are configured on a network interface card. In this regard, the method 300 further comprises controlling an operation of the analog cancellator 206 and the digital cancellator 208 based on a control signal received by the analog cancellator 206 and the digital cancellator 208, respectively.

[0033] In various first embodiments, the network interface card is configured to be transitionable amongst a plurality of states based on a type of a message received or sent. The control signal for the analog cancellator 206 and the digital cancellator 208 corresponds to the message received or sent.

[0034] In various first embodiments, the plurality of states comprise a communication state, a monostatic sensing state and a bistatic sensing state. In this regard, the network interface card is configured to be in the monostatic sensing state, and the analog cancellator 206 and the digital cancellator 208 are controlled to be enabled based on the message indicating data transmission or acknowledgement of a predetermined type of data reception.

[0035] In various second embodiments, the sensing system 200 comprising a sensing device is removably connected to a Wi-Fi network interface card. In this regard, the sensing device comprises the analog cancellator 206, the digital cancellator 208 and the channel state information generator 216.

[0036] In various second embodiments, the sensing device is a software-defined radio (SDR).

[0037] In various embodiments, as shown in FIG. 2, the sensing system 200 comprises: at least one memory 202; and at least one processor 204 communicatively coupled to the at least one memory 202 and configured to perform various steps or operations of the above-mentioned method 300 of monostatic sensing over Wi-Fi communication as described hereinbefore according to various embodiments of the present invention. In various embodiments, the at least one processor 204 is configured to: receive (at 306) (performed by the digital cancellator 208) the transmission signal suppressed analog signal (from the analog cancellator 206); suppress (at 308) (performed by the digital cancellator 208) the residue transmission signal component to produce a transmission signal suppressed digital signal; and generate (at 310) (performed by the channel state information generator 216) channel state information for monostatic sensing based on the transmission signal suppressed digital signal.

[0038] It will be appreciated by a person skilled in the art that the at least one processor 204 may be configured to perform various functions or operations through set(s) of instructions (e.g., software modules) executable by the at least one processor 204 to perform various functions or operations. Accordingly, as shown in FIG. 3, the sensing system 200 may comprise the digital cancellator 208 configured to receive the transmission signal suppressed analog signal and to suppress the residue transmission signal component for producing a transmission signal suppressed digital signal; and the channel state information generator 216 configured to generate channel state information for monostatic sensing based on the transmission signal suppressed digital signal.

[0039] It will be appreciated by a person skilled in the art that the digital cancellator 208 and the channel state information generator 216 are not necessarily separate modules, and two or more modules may be realized by or implemented as one functional module (e.g., a software program) as desired or as appropriate without deviating from the scope of the present invention. For example, the digital cancellator 208 and the channel state information generator 216 may be realized (e.g., compiled together) as one executable software program (e.g., software application or simply referred to as an “app”), which for example may be stored in the at least one memory 202 and executable by the at least one processor 204 to perform various functions/operations as described herein according to various embodiments of the present invention.

[0040] For example, in various embodiments, the at least one memory 202 may have stored therein the digital cancellator 208 and the channel state information generator 216, which respectively correspond to various steps (or operations or functions) of the method 300 of monostatic sensing over Wi-Fi communication as described herein according to various embodiments, which are executable by the at least one processor 204 to perform the corresponding functions or operations as described herein. For example, the digital cancellator 208 and the channel state information generator 216 may be implemented in a universal software radio peripheral (USRP).

[0041] A computing system, a controller, a microcontroller or any other system providing a processing capability may be provided according to various embodiments of the present invention. Such a system may be taken to include one or more processors and one or more computer-readable storage mediums. For example, as described hereinbefore, the sensing system 200 for monostatic sensing over Wi-Fi communication may include at least one processor (or controller) 204 and at least one computer-readable storage medium (or memory) 202 which are for example used in various processing carried out therein as described herein. A memory or computer-readable storage medium used in various embodiments may be a volatile memory, for example a DRAM (Dynamic Random Access Memory) or a non-volatile memory, for example a PROM (Programmable Read Only Memory), an EPROM (Erasable PROM), EEPROM (Electrically Erasable PROM), or a flash memory, e.g., a floating gate memory, a charge trapping memory, an MRAM (Magnetoresistive Random Access Memory) or a PCRAM (Phase Change Random Access Memory).

[0042] In various embodiments, the radio frequency switch 210 and the analog cancellator 206 may each be implemented as a circuit. For example, a circuit may be understood as any kind of a logic implementing entity, including a special purpose circuitry. Accordingly, the sensing system 200 may be implemented as a combination of hardware module(s) (functional hardware unit(s)/circuit(s) designed to perform the required functions or operations, such as the radio frequency switch 210 and the analog cancellator 206) and software module(s) (computer program(s) or set(s) of instructions executable by a computer processor to perform the required functions or operations, such as the digital cancellator 208 and the channel state information generator 216).

[0043] FIG. 4 depicts a schematic flow diagram of a method 400 of forming a sensing system for monostatic sensing over Wi-Fi communication, such as the sensing system 200 as described hereinbefore according to various embodiments of the present invention. The method 400 comprising: providing (at 402) an analog cancellator 206 configured to receive an input analog signal comprising a transmission signal component and a plurality of multi-path reflection signal components and to suppress the transmission signal component of the input analog signal for producing a transmission signal suppressed analog signal comprising a residue transmission signal component; providing (at 404) a digital cancellator 208 configured to receive the transmission signal suppressed analog signal and to suppress the residue transmission signal component for producing a transmission signal suppressed digital signal; providing (at 406) a radio frequency switch 210 configured to selectively switch a coupling to a communication port from between an antenna 212 and a predetermined load 214 for facilitating optimization of a parameter associated with the analog cancellator 206 for suppressing the transmission signal component of the input analog signal and a parameter associated with the digital cancellator 208 for suppressing the residue transmission signal component; and providing (at 408) a channel state information generator 216 configured to generate channel state information for monostatic sensing based on the transmission signal suppressed digital signal.

[0044] In various embodiments, the method 400 is for forming the sensing system 200 as described hereinbefore according to various embodiments, therefore, various steps or operations of the method 400 may correspond to forming, providing or configuring various components or elements of the sensing system 200 as described herein according to various embodiments, and thus such corresponding steps or operations need not be repeated with respect to the method 400 for clarity and conciseness. In other words, various embodiments described herein in context of the sensing system 200 are analogously valid for the method 400 (e.g., for forming the sensing system 200 having various components and configurations as described herein according to various embodiments), and vice versa.

[0045] It will be appreciated by a person skilled in the art that the terminology used herein is for the purpose of describing various embodiments only and is not intended to be limiting of the present invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.

[0046] Any reference to an element or a feature herein using a designation such as “first”, “second” and so forth does not limit the quantity or order of such elements or features, unless stated or the context requires otherwise. For example, such designations may be used herein as a convenient way of distinguishing between two or more elements or instances of an element. Thus, a reference to first and second elements does not necessarily mean that only two elements can be employed, or that the first element must precede the second element. In addition, a phrase referring to “at least one of’ a list of items refers to any single item therein or any combination of two or more items therein.

[0047] In order that the present invention may be readily understood and put into practical effect, various example embodiments of the present invention will be described hereinafter by way of examples only and not limitations. It will be appreciated by a person skilled in the art that the present invention may, however, be embodied in various different forms or configurations and should not be construed as limited to the example embodiments set forth hereinafter. Rather, these example embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the present invention to those skilled in the art.

[0048] Although Wi-Fi communications have been exploited for sensing purpose for over a decade, the bistatic or multistatic nature of Wi-Fi still poses multiple challenges, hampering real-life or practical systems. In this regard, various example embodiments seek to re-design Wi-Fi so that monostatic sensing (mimicking radar) can be achieved over the multistatic communication infrastructure. In particular, various example embodiments provide integrated sensing and communication (ISAC) within the Wi-Fi framework, which may be herein referred to as ISAC-Fi. Various example embodiments propose, design and implement ISAC-Fi as an ISAC-ready Wi-Fi system or device, including example prototypes. In this regard, various example embodiments provide a self-interference cancellation scheme or technique, in order to extract reflected RF signals for sensing purpose in the face of concurrent transmissions. Various example embodiments also revise existing Wi-Fi framework so as to seamlessly operate monostatic sensing under Wi-Fi communication standard. As example implementations, various example embodiments provide two ISAC-Fi designs or configurations, namely, a full version which may be USRP -based that corresponds to a re-designed ISAC-Fi system or device capable of performing integrated sensing and communication, and a partial version which may have a plug-and-play design or configuration that allows for backward compatibility by attaching the plug-and-play device or module (which may be referred to herein as a sensing device or module) to an arbitrary Wi-Fi NIC (e.g., an existing or a commercially available WiFi NIC). As will be described later below according to various example embodiments, extensive experiments were performed to validate the efficacy of ISAC-Fi according to various example embodiments and also to demonstrate its superiority over existing Wi-Fi sensing techniques.

[0049] As described in the background, among all constraints imposed by Wi-Fi’s bistatic or multistatic nature, three prominent constraints are illustrated in FIG. 1 by way of examples only and without limitations. In particular, FIG. 1 depicts a schematic drawing illustrating conventional Wi-Fi sensing, whereby the thin arrows represent various RF propagations originating from the transmitter (Tx) along various distinct paths, while the thick (double-sided) arrows denote the directions of sensed subject motions.

[0050] Various example embodiments note that all the constraints imposed by Wi-Fi’s bistatic or multistatic nature and illustrated in FIG. 1 can be lifted if the sensing mode is able to be converted to be monostatic, whereby the antenna(s) of each Wi-Fi device, while transmitting data packets, also capture the reflected signals induced by the transmissions and certain reflecting subjects, as shown in FIG. 5. In particular, FIG. 5 depicts a schematic drawing illustrating ISAC-Fi sensing according to various example embodiments of the present invention, whereby the thin arrows represent various RF propagations originating from different transmitters (Tx) along various distinct paths, while the thick (double-sided) arrows denote the directions of sensed subject motions. For example, conventional device-based WiFi sensing is a special type of bistatic sensing that aims to locate the Tx. Therefore, the ISAC revision to Wi-Fi according to various example embodiments is orthogonal to this type of sensing applications dedicated solely to localization.

[0051] According to the ISAC-Fi sensing as shown in FIG. 5, the ToFs of these reflection paths can be readily obtained as all uncertainties are removed thanks to the co-location of the Tx and the Rx. Moreover, the AoA of a reflection path can be more accurately estimated without the LoS path interference, exploiting the MIMO (multiple-input and multiple-output) capability of a Wi-Fi device (i.e., its antenna array). Given both ToF and AoA, locating a reflecting subject can thus be achieved by only one device. According to various example embodiments, the localization accuracy can be further improved by leveraging the interaction between a pair of communicating devices (a distributed MEMO setting). In addition, various example embodiments note that operating individual Wi-Fi devices in monostatic mode advantageously maximizes the utilization of radio frequency resources, as opposed to the serious wastes under the bistatic setting, because two devices in the latter setting obtain far less information than only one device in the former. Last but not least, the interpretation of any motion effect sensed over a reflection path is clearly defined without any ambiguity.

[0052] However, ISAC within Wi-Fi framework by performing sensing in monostatic mode is far from straightforward. In particular, a major technical challenge or problem is that the cost of removing LoS path interference is the self-interference from Tx to (its own) Rx. For example, normal radars may rely on ultra-wide bandwidth (hence nanosecond time resolution) to separate this Tx-interference, which may not be available to Wi-Fi in the current or the next few generations.

[0053] Accordingly, various example embodiments provide monostatic sensing over Wi-Fi communication, including a sensing system and a method for monostatic sensing over Wi-Fi communication, that seek to overcome, or at least ameliorate, one or more deficiencies or constraints in bistatic or multistatic sensing over Wi-Fi communication, and more particularly, in an effective and efficient manner. In particular, various example embodiments advantageously provide ISAC-Fi that seeks to enable integrated sensing and communication within the Wi-Fi framework. In this regard, various example embodiments achieve Tx-Rx signal separation within the same RF chain by revising the idea of full-duplex radios so as to operate monostatic sensing over Wi-Fi communications. As example implementations, various example embodiments provide two ISAC-Fi designs or configurations (including two prototypes of ISAC-Fi), namely, a full version 610 (which may be referred to herein as a full ISAC-Fi) which makes use of USRP (e.g., USRP X310) that may be implemented as a Wi-Fi NIC (network interface card) with a front-end capable of cancelling or minimizing the Tx- interference such as shown in FIG. 6B (depicting a schematic drawing of an example architecture of the full ISAC-Fi 610); and a partial version (which may be referred to herein as a partial ISAC-FI) which applies a plug-and-play (PnP) module (sensing device or module) 620 to an arbitrary Wi-Fi NIC 600 (e.g., an existing or a commercially available Wi-Fi NIC) such as shown in FIG. 6C (depicting a schematic drawing of an example architecture of the partial ISAC-Fi 630), delivering a backward compatibility while rendering conventional Wi-Fi NICs 600 ISAC -ready. For comparison, FIG. 6A depicts a schematic drawing of an example architecture of a conventional Wi-Fi NIC 600.

[0054] For example, compared to the conventional Wi-FI NIC 600 shown in FIG. 6A, in the example full ISAC-Fi version 610 shown in FIG. 6B, the Tx-Rx switch 602 may be replaced with a Tx-Rx signal separator 612 according to various example embodiments of the present invention. This enables the Tx and Rx to operate simultaneously (instead of in a time-divided manner in the conventional Wi-FI NIC 600 shown in FIG. 6A). Therefore, while transmitting signal by the Tx, the Rx can concurrently receive a signal, which is a combination of the Tx signal (partially, as undesirable interference) and the multi-path reflection signals of the Tx signal (which may herein be referred to as a Tx-Rx signal), based on which monostatic sensing is performed according to various example embodiments of the present invention. The partial ISAC-Fi version 630 seeks to achieve monostatic sensing in the same or similar manner as the full ISAC-Fi version 610 show in FIG. 6B, but with an unaltered Wi-Fi NIC 600 (e.g., an existing or a commercially available Wi-Fi NIC, such as the conventional Wi-Fi NIC 600 shown in FIG. 6A). Therefore, the conventional Wi-Fi NIC 600 may operate as originally configured (e.g., default communication functions), while the PnP module 620 is configured to perform monostatic sensing function according to various example embodiments of the present invention. Both the PnP module 620 and the Wi-Fi NIC 600 share the same antenna and operate simultaneously, whereby the Tx-Rx signal is separated by the Tx-Rx signal separator 612. Furthermore, the PnP module 620 and the Wi-Fi NIC 600 are synchronized according to various example embodiments of the present invention.

[0055] In various example embodiments, existing Wi-Fi MAC protocol may be fine-tuned so that both individual and distributed sensing are fully operational without affecting conventional Wi-Fi communications. In this regard, though preserving the existing Wi-Fi MAC protocol may be desiredfor the sake of compatibility, minor tuning of protocol details may be advantageous in, for example, enabling and disabling Tx interference signal cancellation, according to various example embodiments of the present invention.

[0056] Accordingly, various example embodiments advantageously provide a number of contributions such as but not limited to:

• a Wi-Fi based ISAC prototype (ISAC-Fi) compatible with the Wi-Fi framework;

• a new RF front-end to replace that of the conventional Wi-Fi design for effectively separating concurrent sensing and communication signals (Tx-Rx signals);

• an example full prototype of ISAC-Fi leveraging the universal emulation capability of USRP (e.g., USRP X310); and

• an example partial prototype of ISAC-Fi which provides a PnP module to a conventional or existing Wi-Fi NIC so as to elevate it to be ISAC -ready.

ANALYSIS OF EXISTING WI-FI SENSING

[0057] For better understanding, basic theoretical and experiment analyses are provided to compare bistatic Wi-Fi sensing with monostatic mode according to various example embodiments in terms of channel model. For example, these analyses and comparisons demonstrate the weaknesses of existing bistatic Wi-Fi sensing and provide a background for the design or development of the ISAC-Fi according to various example embodiments of the present invention. In general, the Wi-Fi OFDM (orthogonal frequency-division multiplexing) signal x(t, r) received over the air and modulated onto a certain carrier frequency f _c is given by:

(Equation 1) denotes h(t, z) (channel over the air), e ^{-727T ct} s(t) denotes the Tx baseband symbol s(t) after up-conversion, symbol * denotes convolution, T _p and Tp (t) denote the propagation delay and the motion-induced delay along the -th propagation path, respectively. For example, these are key sensing information offered by Wi-Fi communications. Uncertainties in Temporal Features

[0058] According to various example embodiments, uncertainties of the temporal features are characterized in a channel model, and their implications are also discussed.

[0059] Modelling Offsets. Since the crystal oscillators (i.e., clocks) of Tx and Rx may differ slightly, the resulting imperfect signal processing introduces several random offsets to contaminate both To understand the details of these errors, the whole processing line of an example Rx chain will now be described. First of all, down-converting the OFDM signal x(t, r) in the Rx chain requires applying to shift x(t, T) to baseband, but the resulting baseband signal is actually:

(Equation 2) where y _c = f _c — _c ' denotes the CFO (Carrier Frequency Offset) caused by the residue error in PLL (Phase Locked Loop); it forces Rx to match f _c with a slightly different _c '.

[0060] Moreover, a CPO (Carrier Phase Offset) <p _c is imposed by both PLL and VCO (Voltage Controlled Oscillator) since VCO has a random phase each time it starts or restarts and PLL cannot fully compensate the phase difference between the local Rx carrier and the received signals x(t, T).

[0061] Further down the processing line of the Rx chain, the baseband signal y(t, r) is sampled by ADC (analog-to-digital converter) and then converted to frequency domain via FFT (fast Fourier transform). Considering an OFDM symbol with size AFT and Rx sampling period T _s = ff ¹ with f _s being the sampling rate, various example embodiments let t = nT _s with n denoting the sampling index and thus obtain the following fc-th sub-carrier signal of the /-th OFDM symbol after FFT : (Equation 3) where H(k, T) = is the OFDM sub-carrier spacing, and with T _s ' denoting the Tx sampling period, is the SFO (Sampling Frequency Offset) caused by the difference between Tx DAC (digital-to-analog converter) and Rx ADC clocks.

[0062] Finally, due to the lack of knowledge on the starting point of an OFDM symbol at the Rx side, it is hard to determine the right samples to feed into FFT. This issue persists even with a carefully designed preamble and corresponding detection algorithms, causing a phase error because missing even a small length of the preamble equivalently results in a non- negligible delay. Various example embodiments term this phase error PDD (Packet Detection Delay) e; and thus, Y _f (k, T) in Equation (3) may be revised to:

(Equation 4) [0063] Although all these errors exist in normal Wi-Fi communications, they have been masked by well-designed demodulation schemes. However, sensing aims to capture minor variations, rendering it intolerable to even minor errors and hence fundamentally different from communication.

[0064] Bistatic vs. Monostatic Sensing. All uncertainties in the channel model in Equation (4) (i.e., CFO, CPO, SFO, and PDD) may affect bistatic sensing. Therefore, it is extremely challenging (if it is even possible) to measure quantities induced by temporal features (e.g., ToF from T _p ). On the contrary, by switching to the monostatic mode so that Tx and Rx become colocated in the same device according to various example embodiments of the present invention, they would share the same clock. Therefore, CFO, SFO, and PDD can be significantly reduced. Although CPO may still persist, obtaining it during the hardware initialization is viable according to various example embodiments of the present invention. For example, the CSI phases of the same symbol in consecutive packets were measured under both bistatic and monostatic modes in an empty room. FIGs. 7A and 7B show plots of the unwrapped CSI phases of 52 subcarriers and across consecutive symbols under bistatic mode and monostatic mode, respectively. As shown in FIGs. 7A and 7B, the phases under the bistatic mode increase gradually with I (symbol) and have smaller slopes in k (subcarrier) than those under the monostatic mode, which accords well with the phase terms in Equation (4) given negative [J and e. The monostatic mode, on the contrary, exhibits only minor phase variations across both and consistent slops in k, thus allowing for accurate recovery of temporal features z _p and Tp (t). The glitches at the O-th subcarrier in FIG. 7A are caused by the lack of data stream and the phase unwrapping process, which may jump drastically due to CFO under the bistatic mode but are well controlled otherwise.

Dominating Interference from LoS Path

[0065] According to the channel model in Equation (1), multiple signal propagation paths exist, among them two are special: the O-th path (corresponding to the Tx-interference, which will be described further later below according to various example embodiments of the present invention) and the 1-st path (the LoS path between Tx and Rx) which has a dominating power over all other NLoS paths under the bistatic mode but it disappears under the monostatic mode. To better understand the LoS path interference to NLoS paths, the amplitude a _P in Equation (1) is expanded to calculate the received power P _p ^Rx at an Rx antenna as follows:

(Equation 5) where P ^Tx is the Tx power, G ^Tx and G ^Rx are the Tx and Rx antenna gains, A _c denotes the wavelength of the carrier frequency, <J _P represents the RCS (Radar Cross Section) of the reflecting target, and /?J ^X and /? ^Rx represent the Tx-target and target-Rx ranges. As far as the target does not lie on the LoS path (which is very rare), the power ratio between the p-th (NLoS) path signal carrying the (reflected) sensing information and the interfering LoS path signal becomes:

(Equation 6) where L denotes the LoS distance between Tx and Rx.

[0066] Assuming a bistatic sensing with L = 2 m, <J _P = 1 m ² , and R _p ^x = /? ^Rx = 2 m just for simplicity and without limitation, Equation (6) suggests that r/ _p « —40 dB. It should be noted that as the LoS path signal is meant for communications (the main function of Wi-Fi), it cannot be suppressed just for sensing purpose. In contrast, according to various example embodiments, operating the sensing function under the monostatic mode could totally remove the LoS path; in other words, the constraint imposed by Equation (6) disappears. In fact, under the monostatic mode, the Tx-target-Rx roundtrip path becomes the dominating one (e.g., see FIG. 5), and it advantageously carries the desired sensing information according to various example embodiments of the present invention. For example, a set of experiments were performed to briefly demonstrate the differences, where L = 1.5 m was set and the target-Rx ranges were varied. FIGs. 8A and 8B depict phase variations induced by human (target) breath at two different Rx-target ranges under bistatic mode and monostatic mode, respectively. As shown in FIGs. 8A and 8B, the LoS interference is very evident under the bistatic mode, especially when compared with those under the monostatic mode. However, various example embodiments note a new challenge, or technical problem, under the monostatic mode: the O-th path (or Tx- interference) signal in Equation (1) (absent under the bistatic mode due to the temporal separation enforced by CSMA/CA MAC protocol) will cause a serious problem. Accordingly, various example embodiments seek to overcome, or at least ameliorate, this technical problem associated with the monostatic mode so as to enable, or facilitate, monostatic sensing over WiFi communication.

Ambiguity in Motion Sensing

[0067] The motion-induced delay T _p in Equation (1) will now be discussed, which is a quantity representing the variations (e.g., target motion) along the p-th path. Basically, T _p (t) = AR„(t)

— — where c is the speed of light and A/? _p (t) is the instantaneous variation in range at time t. Though A/? _p (t) is often termed displacement in radar terminology, it is actually a scalar obtained by projecting the actual displacement of a moving target onto a certain direction. Although this direction is concretely determined under the monostatic mode (as it is exactly the Tx/Rx-target direction shown in FIG. 5), it can be highly ambiguous under the bistatic mode.

[0068] Because A/? _p (t) represents the variation in range and the range is actually the length of Tx-target-Rx reflection path under the bistatic mode, a field with Tx and Rx can be defined as two focus points. As this specific field describes the lengths of Tx-target-Rx reflection paths, its equipotential surfaces correspond to equal-length contours that happen to be ellipsoids with Tx and Rx as foci (see FIG. 9). In particular, FIG. 9 shows the conceptual illustration and experiment settings for the sensing motion effect under bistatic mode. At any point in the field, a target displacement d can be decomposed into tangent and normal components based on the ellipsoid on which the target resides. Since A/? _p (t) senses the variant in range, it can only represent the normal component whose direction varies with the target location, whereas the tangent component leads to variation along an equal-length contour and thus delivers no impact on A/? _p (t) . Further reasoning could deduce that all normal directions lie on hyperbolas confocal with (hence orthogonal to) the ellipsoids, which can be deemed as the field lines of this field.

[0069] Although it is highly non-trivial to experimentally characterize this field, verifying the ambiguity in sensing motion direction can be readily obtained by well-designed experiments. For example, various example embodiments adopt a motor-driving slide rail that is programmable to move a target in a constant speed. According to the aforementioned analysis, putting the rail parallel or perpendicular to the Tx-Rx line (as shown by the thick double-side arrows in FIG. 9) and varying its position within the field, the sensed A/? _p (t) should exhibit magnitude variations even though the target is programmed to have a constant speed along the rail, simply due to the monotonically varying projections onto the field lines. FIGs. 10A and 10B show plots of the unwrapped CSI phases of the 1st subcarrier when a motor-driving slide rail is placed perpendicular (FIG. 10A) and parallel (FIG. 10B) to the Tx-Rx line. As shown in FIGs. 10A and 10B, the monotonic trends of A/? _p (t) (represented by phases) are evident under both perpendicular and parallel cases, firmly corroborating the analysis discussed above.

[0070] It is worth noting that enabling the monostatic mode in Wi-Fi according to various example embodiments does not replace its originally bistatic sensing ability, because communications are still half-duplex as defined by CSMA/CA MAC. Instead, it simply exerts the full potential of Wi-Fi sensing over communications. In particular, instead of having a pair of Wi-Fi devices working as one bistatic radar, various example embodiments can simultaneously have two monostatic and one bistatic radars. Based on this ISAC architecture on Wi-Fi, various example embodiments perform fine-tuning, or modification, of the MAC protocol to differentiate the Rx status under different radar modes.

ISAC-Fi: MAKING WI-FI ISAC-READY

[0071] Design of the RF front-end and two example architectures or prototypes (namely, the full and partial architectures or prototypes) of ISAC-Fi will be described below according to various example embodiments of the present invention.

[0072] An example design of ISAC-Fi will now be described according to various example embodiments of the present invention. In this regard, various example embodiments first explain how to realize the Tx-Rx signal separator (e.g., corresponding to the Tx-Rx signal separator 612 shown in FIGs. 6B and 6C) as the basic enabler of ISAC-Fi (for both the full and partial prototypes). Subsequently, various example embodiments discuss the potential issues and corresponding countermeasures for both co-existence with Wi-Fi framework and channel parameter estimation under irregular traffic. In addition, an example implementation of collaborative MIMO (multiple-input multiple-output) sensing under ISAC-Fi will also be described according to various example embodiments of the present invention.

System Overview

[0073] In various example embodiments, the design of ISAC-Fi may be centered around the ability of separating, or configured to separate, concurrent Tx and Rx signals (which may be referred to as Tx-Rx signals herein), and more particularly, the Tx-Rx signal separator. FIG. 11 A depicts a schematic drawing of an example first or full version of ISAC-Fi 1100 (or simply referred to as the first or full ISAC-Fi) with Tx-Rx signal separation according to various example embodiments of the present invention. FIG. 1 IB depicts a schematic drawing of an example second or partial version of ISAC-Fi 1150 (or simply referred to as the second or partial ISAC-Fi) with Tx-Rx signal separation according to various example embodiments of the present invention.

[0074] In various example embodiments, the full version of ISAC-Fi 1100 is configured to integrate both sensing and communication functions. However, as the full version of ISAC-Fi 1100 requires revamping or modifying the design of Wi-Fi NIC, various example embodiments additionally provide a partial version of ISAC-Fi 1150 with backward compatibility to an existing or a commercially available Wi-Fi NIC 1170 (corresponding to the conventional WiFi NIC 600 as described hereinbefore with reference to FIGs. 6A and 6C). In this regard, the partial ISAC-Fi 1150 comprises a sensing device (which may also be referred to herein as a sensing module) 1160 (corresponding to the PnP module 620 as described hereinbefore with reference to FIG. 6C) appendable or removably connectable to a Wi-Fi NIC 1170 (e.g., an existing or a commercially available Wi-Fi NIC).

[0075] In various example embodiments, the Tx-Rx signal separator according to various example embodiments differs significantly from full-duplex radios in that, for example, the Rx (monostatic sensing) signal to be extracted is a reflection of the corresponding (slightly earlier) Tx signal. In various example embodiments, the full ISAC-Fi 1100 comprises a circulator 1104 at a first stage configured to physically separate Tx and Rx signals (roughly separate, that is, would still have a portion of the Tx signal (i.e., undesirable Tx interference) remaining). In various example embodiments, the partial ISAC-Fi 1150 comprises a hybrid coupler 1154 configured to isolate the Rx signal (roughly isolate, that is, would still have a portion of the Tx signal (i.e., undesirable Tx interference) remaining) before feeding it to the sensing module 1160. In various example embodiments, remaining components for suppressing the residue Tx- interference may be common or the same in the full and partial versions. As will be described later below, the full and partial versions 1100, 1150 may each comprise an analog cancellator 1180 and a digital cancellator 1190 that are configured to preserve the monostatic sensing signal. In various example embodiments, the Tx-Rx signal separator (e.g., corresponding to the Tx-Rx signal separator 612 shown in FIG. 6B) of the full ISAC-Fi 1100 may comprise the circulator 1104, the analog cancellator 1180 and the digital cancellator 1190. In various example embodiments, the Tx-Rx signal separator (e.g., corresponding to the Tx-Rx signal separator 612 shown in FIG. 6C) of the partial ISAC-Fi 1150 may comprise the hybrid coupler 1154, the analog cancellator 1180 and the digital cancellator 1190.

[0076] In various example embodiments, the full ISAC-Fi 1100 is configured to be compatible with the Wi-Fi framework, for example, the analog and digital cancellators 1180, 1190 may be configured to not operate under the reception mode (i.e., bistatic or normal Wi-Fi communication mode). In this regard, various example embodiments leverage the DATA/ACK messages to invoke transitions among three major states of ISAC-Fi, namely communication, monostatic sensing, and bistatic sensing, as will be described further later below under the subheading “Co-existing with Wi-Fi Framework'' according to various example embodiments of the present invention.

[0077] After the Tx-Rx signal separation, in various example embodiments, signal processing is applied to treat bistatic and monostatic signals separately. Since existing Wi-Fi sensing techniques assume regular data packets that are far from realistic, in various example embodiments, the signal processing may be configured to make sensing compatible with irregular data packets, as will be described further later below under the sub-heading “Monostatic Channel Feature Estimation" according to various example embodiments of the present invention.

[0078] Although monostatic sensing with respect to single-device operations of ISAC-Fi may have been described according to various embodiments or example embodiments of the present invention, by sharing information as appropriate provided by both the Wi-Fi and/or backbone (wired) networks, a set of Wi-Fi devices (e.g., AP(s) (access point(s)) and full/partial ISAC-Fi(s)) may be configured to collaboratively serve as a distributed MIMO sensing system according to various example embodiments of the present invention. By way of an example only and without limitation, an example distributed MIMO sensing system will be described later below under the sub-heading “Collaborative MIMO Sensing according to various example embodiments of the present invention.

Tx-Rx Signal Separator Design

[0079] Various example embodiments note that because the monostatic sensing signals (corresponding to the Rx signals) are reflected version of the corresponding Tx signals, they would be treated as interference from the perspective of conventional full-duplex radios. In stark contrast, various example embodiments seek to preserve these monostatic sensing signals while removing the Tx-interference. In this regard, the analog cancellator 1180 and the digital cancellator 1190 of the full ISAC-Fi 1100 and the partial ISAC-Fi 1150 will now be further described according to various example embodiments of the present invention, including improved self-adapted Tx-Rx signal separation.

Analog Cancellator 1180

[0080] In various example embodiments, the analog cancellator 1180 is configured to receive the output of the circulator (in the case of a full ISAC-Fi 1100) or the hybrid coupler (in the case of a partial ISAC-Fi 1150) as its analog input. Since the received signal contains Tx-interference (p = 0) (e.g., corresponding to the “transmission signal component” as described hereinbefore according to various embodiments) and multi-path reflections (p > 0) (e.g., corresponding to the “plurality of multi-path reflection signal components” as described hereinbefore according to various embodiments) via different channels respectively, according to various example embodiments, the output of the analog cancellator 1180 may be expressed as:

Equation (7) where G _A , G _H , and G _c denote the channel gains of the analog cancellator 1180, the RF hardware (e.g., all hardware components which the Tx-interference signal passes through), and the circulator 1104 or hybrid coupler 1154, respectively. In various example embodiments, T _p denotes both T _p and T _p in Equation (1). Let the residue Tx-interference (e.g., corresponding to the “residue transmission signal component” as described hereinbefore according to various embodiments) be m _A (t) = G _A ■ x(t, T ₀ ) + G _H ■ x(t, T ₀ ) . In this regard, according to various example embodiments, the analog cancellator 1180 is configured to adjust G _A so as to minimize m _A (t). In various example embodiments, a Direct Quadrature Modulator (DQM) is provided or configured to realize the analog cancellator 1180, and an architecture or a configuration of an example DQM 1206 is shown in FIG. 12. In particular, FIG. 12 depicts a schematic drawing of the analog cancellator 1180 implemented in the form of a DQM 1206 and communicatively coupled to the circulator 1104 (in the case of a full ISAC-Fi 1100) or the hybrid coupler 1154 (in the case of a partial ISAC-Fi 1150). For example, implementing the analog cancellator 1180 in the form of a DQM 1206 advantageously provides a much simpler yet more effective architecture that treats G _A as the inverse of G _H , and more particularly, it controls the IQ baseband generator of the DQM 1206 to match G _A with G _H in order to minimize the residue Tx-interference m _A (t). For example, the DQM 1206 is a feedback control system driven by the residue Tx-interference m _A (t) , whereby as long as m _A (t) is not zero, a control signal is generated to excite both DACs on the I/Q components of RF signal. The outcome of this control loop is to produce, at the output of the DQM 1206, a signal that is configured to cancel the Tx- interference. For example, employing the DQM 1206 has been advantageously found to be more compact and effective (e.g., with wider dynamic range and higher resolutions in both amplitude and phase) than a fixed delay circuit.

Digital Cancellator 1190

[0081] Various example embodiments note that since ISAC-Fi demands only CSI extraction for sensing but has no interest on data contents, a preamble-based digital cancellator 1190 is provided according to various example embodiments. In this regard, the digital cancellator 1190 is configured to sample the output preambles from the analog cancellator 1180 via correlations, which have passed through the RF downconversion and ADC sampling, thus containing C _A nT _s ) as the residue analog Tx-interference:

Equation (8) where G _D denotes an adaptive filter whose coefficients are obtained via the Least Mean Squares (LMS) algorithm with low complexity and fast convergence. As a result, a linear combination of multiple time-delayed versions of s(nT _s ~) is constructed by applying G _D , aiming to offset m (nT _s ). In this regard, s(nT _s ~) represents the baseband of the known Wi-Fi preamble prestored by ISAC-Fi, and using it avoids the much larger errors in LMS processing inherent to the data part.

Self-Adapted Tx-Rx Signal Separation

[0082] In relation to the analog cancellator 1180 and the digital cancellator 1190 described above, various example embodiments note that the adjustments to G _A and G _D may not perfectly focus on the Tx-interference. In various example embodiments, the minimization of the Tx- interference is performed by the analog cancellator 1180 and the digital cancellator 1190 in sequence, and hence, the minimization may be applied to z(t, r) and z(nT _s , T) one at a time (i.e., one after another). In this regard, various example embodiments note that the minimization of the Tx-interference by one cancellator may thus potentially offset the third term of Equation (7) or (8) (i.e., x(t, associated with the other cancellator. However, various example embodiments note that the CSIs contained in these terms (x(t, and are valuable information required by monostatic sensing. Accordingly, various example embodiments advantageously address this technical challenge or problem of how to keep x(nT _s , while removing or minimising the Tx-interference. For better understanding, this technical challenge is illustrated using an experiment of sitting a human subject close to ISAC-Fi (with the analog cancellator 1180 and the digital cancellator 1190 but without the selfadapted Tx-Rx signal separation) as shown in FIG. 13. In particular, FIG. 13 illustrates the erasing of the monostatic sensing signal by the analog and digital cancellators 1180, 1190 (without the self-adapted Tx-Rx signal separation). In FIG. 13, the breath signals have been artificially amplified to be more conspicuous. According to FIG. 13, while applying the analog and digital cancellators 1180, 1190 (without the self-adapted Tx-Rx signal separation) (before 8s with the spikes indicating preamble receptions and hence cancellator recalibrations) which suppresses the Tx-interference below the noise floor, the human breath signal captured by x(nT _s , also disappears. On the contrary, as shown in FIG. 13, stopping the cancellations bring back both the Tx-interference and the breath signal, albeit the latter being heavily distorted by the former.

[0083] Various example embodiments advantageously made certain observations resulting in a technical solution. As a first observation, various example embodiments observe that almost the entire Tx-interference x(t, T ₀ ) comes from hardware circuits rather than the antenna. In this regard, two experiments were performed where RF absorbing materials were used to surround the antenna in one case, and the antenna was replaced by an RF dummy load (or a predetermined load) to stop radiating RF signals in another case. The Tx-interferences under these two cases were compared and shown in FIG. 14 A. FIG. 14A shows a plot of the correlation coefficients of the Tx-interferences under the above-mentioned two antenna settings. As can been seen, FIG. 14A shows that the correlation cofficients between them are over 90%, demonstrating that the antenna has almost no impact on the Tx-interference. As a second observation, various example embodiments observe that the hardware channel of Tx- interference is stable in a long term. Therefore, with proper calibrations on both G _A and G _D (e.g., according to Equations (7) and (8)), they can keep cancelling the Tx-interference without further fine-tuning in at least ten minutes, as shown in FIG. 14B. FIG. 14B shows a plot of the correlation coefficients of the Rx signal right after calibrations and those received later.

[0084] In view of the above observations, according to various example embodiments, an RF switch 1210 is provided or configured to toggle between an antenna 1212 and a dummy load 1214 (e.g., corresponding to the “predetermined load” as described hereinbefore according to various embodiments) in the ISAC-Fi design, so as to realize a self-adapted Tx-Rx signal separation. Advantageously, this requires only a minor variation bearing negligible complexity and monetary costs. In various example embodiments, the RF switch 1210 may be provided or configured as shown in FIG. 12. Basically, ISAC-Fi switches its Tx port from the antenna 1212 to the dummy load 1214 in a regular basis or before enabling monostatic sensing, allowing for properly calibrating both G _A and G _D (e.g., according to Equations (7) and (8)). For example, during communications and monostatic sensing, ISAC-Fi switches its Tx port to the antenna 1212 and leverages both the analog and digital cancellators 1180, 1190 with G _A and G _D to suppress the Tx-interference. In this regard, switching to the dummy load 1214 aims to calibrate GA and GD (e.g., to match the Tx-interference to the Rx signal) via their respective optimization processes under a “clean” environment without multi-path reflections. FIG. 15 shows a plot of the human breath signals with and without SA (self-adapted Tx-Rx signal separation). As shown in FIG. 15, the human breath signal is clearly extracted with the selfadapted Tx-Rx signal separation according to various example embodiments of the present invention; otherwise the extraction of the human breath signal can be negatively affected by the residual Tx-interference as a result of incomplete cancellation by the analog and digital cancellators without self-adapted Tx-Rx signal separation.

Co-existing with Wi-Fi Framework

[0085] Accordingly, the above successfully implemented Tx-Rx signal separator according to various example embodiments can readily enable monostatic sensing on any Wi-Fi NICs in general. To improve compatibility with existing Wi-Fi protocol framework, various example embodiments note that if the separator treats all incoming signals indiscriminately, it may strongly affect normal Wi-Fi communications due to its filtering (thus distortion) on Rx signals that potentially affects the demodulation performance. For example, as shown in FIGs. 16A and 16B, applying the Tx-Rx signal separator during normal receptions for normal Wi-Fi communications significantly reduces SNR and in turn throughput. In particular, FIGs. 16A and 16B show plots illustrating the self-adapted Tx-Rx signal separator (SA) heavily degrading normal Wi-Fi packet reception quality in terms of both SNR (FIG. 16 A) and throughput (FIG. 16B). Furthermore, it is a waste of computing resource to apply the Tx-Rx signal separator on normal Rx signals for normal Wi-Fi communications. Note that the above-mentioned problem may only apply to the full version of ISAC-Fi 1100, as the partial version of ISAC-Fi 1150 uses a standalone module (sensing module) 1160 to include the Tx-Rx signal separator that produces monostatic sensing signals.

[0086] To address the above-mentioned problem, for the full version of ISAC-Fi 1100 (e.g., including its integrated implementation as an ISAC-ready NIC), various example embodiments provide and implement a revision or modification to the existing Wi-Fi MAC protocol. In particular, various example embodiments add a control path 1120 driven by standard Wi-Fi protocol messages 1122, which entails a function call to the digital cancellator 1190 and a hardware interrupt for the analog cancellator 1180. For example, the Wi-Fi NIC (e.g., the full version of ISAC-Fi 1100 implemented as a Wi-Fi NIC) may start with a C-state (i.e., communication state for normal Wi-Fi communications), and the transition to an M-state (i.e., monostatic state for monostatic sensing) may be invoked by a DATA message 1122 containing any Wi-Fi traffic or an ACK message responding to certain data receptions (normal Wi-Fi receptions). For example, the DATA message 1122 may include NDP (Null Data Packet) frame not officially standardized by IEEE 802.11, if sensing is required when no data traffic is available. For example, DATA messages 1122 (for Wi-Fi traffic or ACK) all invoke “SEND” function and hence followed by a subsequent Tx signal. In various example embodiments, this is when monostatic sensing may be performed, and thus, the Wi-Fi NIC may transition to the M-state and trigger/enable the analog and digital cancellators 1180, 1190 based on such a DATA message 1122 (e.g., corresponding to a control signal) to process the current Tx-Rx signal received. . In various example embodiments, a transition back from the M-state to the C- state may be controlled by a timer fine-tunable as desired or as appropriate, such as to suit surrounding environments. In various example embodiments, a transition from the C-state to the B-state (i.e., bistatic state for bistatic sensing) may be invoked by a reception of either an ACK message or a DATA message 1122 from another Wi-Fi NIC. This is already implicitly assumed by existing Wi-Fi sensing techniques and requires no particular modification to Wi-Fi protocols.

Monostatic Channel Feature Estimation

[0087] Different from existing Wi-Fi sensing techniques hacking Wi-Fi NICs for pure sensing purposes, ISAC-Fi according to various example embodiments advantageously does not operate in such a brute force manner, as it is configured to stay compatible with existing Wi-Fi standard. In this regard, various example embodiments note that the sensing information that piggybacks on data packets (for both monostatic and bistatic) often arrive irregularly due to the inherent nature of Wi-Fi data traffic, which may render existing channel feature estimation techniques invalid.

[0088] As an example experiment, an instance of human slowly walking indoors was performed to illustrate how irregular packets heavily affect channel feature estimation and the resulting heatmaps are shown in FIGs. 17A and 17B. In particular, FIGs. 17A and 17B show the STFT (short-time Fourier transform) heatmaps of a human slowly walking under regular (FIG. 17 A) and irregular packets (FIG. 17B). Conventionally, the motion-induced delay Tp (t) can be estimated using STFT (Short-Time Fourier Transform). Given sensing information conveyed by regular packets, STFT works well to achieve the heatmap of Tp (t) shown in FIG. 17 A, whereby the high energy concentration ranges from 9 to 15 Hz. However, when applying STFT to sensing information conveyed by irregular packets in practice, the resulting heatmap of Tp (t) becomes that shown in FIG. 17B, whereby the high-energy parts are scattered from 0 to 20 Hz. For example, irregular packets introduce noises and thus large errors to machine learning classifiers for human activity recognition. To address this technical challenge, various example embodiments leverage NFFT (Non-uniform Fast Fourier Transform) and sparse optimization to estimate channel features.

[0089] Although the total number of reflections shown in Equation (1) can be large, various example embodiments note that a few reflections should dominate the rest. For example, only reflections with very significant differences in their delays can be identified under a certain bandwidth. Therefore, the path set is sparse and constrained by delayed versions of the known baseband s(t). Let the Tx times of the irregular packets be T ^tx = [T^, T^ ¹ , ••• ], the vector T = h(TD] denote the channel features to be estimated, and T ¹ (F) represent the inverse-NFFT of matrix ^g _S p _arse optimization problem can be formulated as: minlirili

(Equation 9)

(Equation 10) wherell-Hi and ||- 1| ₂ refer to L ¹ and L ² norms, respectively. For example, various example embodiments may adopt the Alternating Direction Method of Multipliers (ADMM) (e.g., as described in S. Boyd, et al., “Distributed Optimization and Statistical Learning via the Alternating Direction Method of Multipliers”, NOW Publishers Inc., Foundations and Trends in Machine Learning, Vol. 3, Issue 1, Jan 2011, pages 1-122) to solve the above problem. In Equation (10), x refers to the signal after the analog and digital cancellators 1180, 1190.

Collaborative MIMO Sensing

[0090] With every Wi-Fi NIC equipped with the monostatic (thus standalone) sensing capability, a large-scale ISAC system with a much wider coverage can be established, by coordinating a set of widely deployed Wi-Fi APs via the Internet. It will be appreciated by a person skilled in the art that the underlying coordination, sitting at the distributed system level, may be configured as desired or as appropriate. By way of an example only and without limitations, according to various example embodiments, a small set of Wi-Fi NICs co-existing in the same collision domain (with two communicating parties as a special case) is considered. A technique for coordinating them in order to seamlessly leverage their monostatic and bistatic sensing capabilities will then be described according to various example embodiments of the present invention.

[0091] Assuming that Wi-Fi devices within the same collision domain are aware of each other in terms of IDs (MAC addresses) and physical locations, the collaborative MIMO sensing method according to various example embodiments may require every one of them to periodically share their sensing information using broadcast. For example, this is a reasonable assumption since sensing information may become meaningless without these baselines and a database containing such information can be preset when deploying each Wi-Fi device. In this regard, the sensing information may refer to either individual estimation results or (compressed) raw CSI data. After receiving a sufficient amount of shared sensing information, each device may invoke a fusion algorithm to combine these information into a final estimation result. For example, for a readily deployable fusion method to achieve this goal, a maximum likelihood algorithm popular in the radar community may be adopted.

IMPLEMENTATION & BENCHMARKING

[0092] The individual functionalities of both the full and partial version of ISAC -Fi 1100, 1150 will be validated below according to various example embodiments, such as by evaluating basic functions of ISAC-Fi and comparing them with existing Wi-Fi sensing proposals. Implementation and Experiment Setup

[0093] According to various example embodiments, prototypes of the full and partial versions of IS AC -Fi 1100, 1150 may be implemented as follows. A circuit board is constructed for the analog cancellator 1180 for both the full and partial ISAC-Fi 1100, 1150. The digital cancellator 1190, control schedules, as well as various sensing algorithms, are implemented in an SDR (Software-Defined Radio) supported by a host PC. By way of examples only and without limitation, the SDR for the full and partial versions 1100, 1150 are USRP X310 and LimeSDR, respectively. All experiments were performed under two scenarios with irregular packets and also other background Wi-Fi traffic, namely, (i) media streaming (UCF101 - Action Recognition Data Set) and (ii) online gaming (StarCraft by BLIZZARD).

Tx-Rx Signal Separator and Common Procedures

[0094] By way of examples only and without limitation, according to various example embodiments, the circulator 1104 is realised by CentricRF CF2040 for the full version 1100 and the hybrid coupler 1154 is realised by TTM X4C25L1-03G for the partial version 1150. In various example embodiments, to process the output of the circulator 1104 or hybrid coupler 1154, for example, the analog cancellator 1180 is designed as a 5 * 5 cm ² Printed Circuit Board made in FR-4. The DQM 1206 is realised by LTC5589, which enables direct modulation of IQ baseband signals at 2.4 GHz carrier frequency and its Serial Peripheral Interface can be used to control the Tx gain, supply current, phase imbalance, and so on. While the digital cancellator 1190 is run by the SDR, various example embodiments further design a General Purpose Input/Output board based on STM32 (an ARM-based MCU) to control the self-adapted Tx-Rx signal separation involving an RF switch (HMC545A). The STM32 is in turn controlled by a host PC via USB. Monostatic sensing algorithms handling irregular packets (as presented above under the sub-heading “Monostatic Channel Feature Estimation"') are implemented in the host PC. According to various example embodiments, the collaborative sensing is realised based on MQTT, a lightweight publish/subscribe messaging protocol for remote devices information exchange.

Digital Processing and Control Protocols

[0095] For the full version 1100, various example embodiments implement the whole WiFi OFDM PHY supporting 20 MHz bandwidth, constellations from BPSK to 64 QAM, and all channel codes (with 1/2, 2/3, 3/4, and 5/6 coding rate). To let monostatic sensing be compatible with CSMA/CA, ISAC-Fi stays normal (the C-state defined hereinbefore under the sub-heading “Co-existing with Wi-Fi Framework''), and leverages the Received Signal Strength Indicator (RSSI) to determine whether a channel is idle. When transmitting data packets, the monostatic sensing (the M-state) is invoked to enable the Tx-Rx signal separator; the transition back to the C-state is triggered by a timer, or the completion of transmission (e.g., maximum Wi-Fi frame duration 5.484ms), whichever is sooner. Bistatic sensing (the B-state) is triggered by packet receptions from another Wi-Fi NIC, and the transition back to the C-state naturally follows the completion of reception.

[0096] For the partial version 1150, the sensing module 1160 is realized by LimeSDR, while the Wi-Fi NIC 1170 is realized as ESP32 (an ARM-based MCU with integrated Wi-Fi) which already offers the Wi-Fi protocol stack. To synchronize LimeSDR and ESP32, an external 40 MHz clock board is designed based on a Temperature Compensated Crystal Oscillator SiT5356. Most state transitions are the same as the full version 1100 except for the trigger for transiting to the M-state. In particular, when the host PC demands the Wi-Fi NIC 1170 to transmit data packets via hardware USB interrupt, it also invokes LimeSDR to the start the Tx-Rx signal separator simultaneously.

Tx-Rx Signal Separation Performance

[0097] The performance and impact of the Tx-Rx signal separation according to various example embodiments will now be discussed. Firstly, the interference cancellation abilities of different components in the Tx-Rx signal separator were quantified. Then, the impact of Tx-Rx signal separation on normal Wi-Fi data traffic was evaluated. The performance of the Tx-Rx signal separation under the video streaming scenario with the Tx power set to 5 dBm was evaluated and the results are shown in FIG. 18. In particular, FIG. 18 shows plots of the power spectrum of the received baseband signal after various components of Tx-Rx signal separators. Since the full version 1100 with circulator 1104 and the partial version 1150 with hybrid coupler 1154 have nearly the same 12 dB cancellation outcome, FIG. 18 only plots the effects of the analog and digital cancellators 1180, 1190 for the full version 1100. It can be observed that the analog and digital cancellators 1180, 1190 further reduce the Tx-interference by 40 dB and 25 dB, respectively. To sum up, the total cancellation was about 77 dB, and the power of residue Tx-interference after Tx-Rx signal separator was very close to the noise floor. [0098] The impact of the Tx-Rx signal separation according to various embodiments on normal Wi-Fi communication will now be discussed, by leveraging UDP -based video streaming and online gaming as the testing scenarios since TCP conceals packet loss. Specifically, the USRP -based full version 1100, as it cannot be configured to operate in a multistatic setting, was evaluated by video streaming, and the partial version was evaluated by online gaming with at least 3 players. Note that these experiment settings were employed to conduct all remaining experiments. Evaluation results of packet delay and packet loss rate are shown in FIGs. 19A and 19B. In particular, FIGs. 19A and 19B show the impact of the Tx-Rx signal separation on normal Wi-Fi communication. The results show that, though ISAC-Fi leverages data packets to perform sensing, it achieves almost the same communication performance as normal Wi-Fi, demonstrating a zero-interference from sensing to communication. Note that in the experiments, the signals were intentionally attenuated by wall blockage to generate discernible results on packet loss; otherwise they are mostly always 0%.

Ranging Performance

[0099] Ranging is a basic yet important function for Wi-Fi sensing, but it can only be achieved under the monostatic mode (as explained hereinbefore under sub-heading “Uncertainties in Temporal Features'"'), whereas existing proposals can only perform rough or relative estimations. To demonstrate the ranging performance of ISAC-Fi according to various example embodiments under irregular data packets using the sensing algorithms introduced hereinbefore under sub-heading “Monostatic Channel Feature Estimation", both IFFT and MUSIC algorithms were chosen as the baselines. In this experiment, a metal cylinder (radius 0.1 m and height 1.2 m) was fixed on a robot car. Controlling the car remotely to move from 1 m to 15 m with a l m step size in a corridor, the ranging errors were obtained and shown in FIGs. 20A and 20B for the full and partial versions 1100, 1150, respectively, for comparison. It was also found that the same algorithm can obtain similar performance on both full and partial versions 1100, 1150. Also, the medians of ranging errors were 1.42 m, 2.84 m, and 4.32 m for ISAC-Fi, MUSIC, and IFFT, respectively, which can mostly be attributed to ISAC-Fi’ s adaptation to irregular data packets.

Motion Sensing Performance

[00100] As discussed hereinbefore under sub-heading “Ambiguity in Motion Sensing", the direction of each sensed motion is concretely defined for ISAC-Fi, namely, the Tx/Rx-target direction. Therefore, various example embodiments are fully able to determine the magnitude and bearing of a motion with at least two ISAC-Fi devices, which can never be truly achieved by the bistatic sensing regardless of how many Wi-Fi NICs are involved. To validate such a performance of ISAC-Fi, the robot car was controlled to move at different speed range from 0.6 m/s to 3.5 m/s in a hall of 20 x 10m ² , and a 10 m spacing was set between two ISAC-Fi devices to measure the velocity using monostatic sensing. In order to obtain the ground truth, two TI 77 GHz millimeter-wave radars were employed to concurrently perform sensing alongside ISAC-Fi.

[00101] FIGs. 21 A and 21B show the the velocity errors of ISAC-Fi and FFT for the full and partial versions 1100, 1150, respectively. The evaluation results shown in FIGs. 21 A and 21B clearly demonstrate that ISAC-Fi (both full and partial versions 1100, 1150) achieves much lower estimation errors than the baseline algorithm leveraging FFT. However, the partial version 1150 seems to perform slightly worse than the full version 1100. Unlike the ranging estimation described hereinbefore under sub-heading “Ranging Performance” relying on individual packets, motion sensing depends on a series of timestamped packets. Therefore, the partial version, compared with the full version 1100, the Tx times of the irregular packets may be not exactly counted due to its relatively casual construction. For example, as introduced hereinbefore under sub-heading “Implementation and experiment Setup”, the triggering signals of the partial version 1150 come from the host PC all the way down to the SDR, passing through application, OS kernel, driver, and hardware, potentially bringing unpredictable temporal uncertainties. In contrast, the integrated design for a true ISAC-Fi implementation (corresponding to the full version 1100) does not have such constraints.

EVALUATION

[00102] Individual functionalities of both the full and partial prototypes 1100, 1150 described hereinbefore according to various example embodiments under heading “Implementation and Benchmarking” will now be further validated and compared with existing Wi-Fi sensing proposals. Although applications of device-free Wi-Fi sensing may be plentiful, they can be roughly classified into three categories, namely localization, activity recognition, and imaging. Therefore, ISAC-Fi’ s performance will be evaluated below based on these categories, while comparing ISAC-Fi with representative proposals for each category where applicable. Device-free Localization

[00103] As one of the key applications of Wi-Fi sensing, device-free localization frees users from holding a Wi-Fi equipped device and solely relies on the deployed Wi-Fi infrastructure to capture the user locations. Nonetheless, existing bistatic sensing cannot fulfill the critical demands raised by this challenging application, mainly due to its incompetence in accurately estimating temporal features (as explained hereinbefore under sub-heading “Uncertainties in Temporal Features’"'). Therefore, mD-Track (the latest bistatic sensing proposal on device-free localization) was chosen as the comparison baseline for demonstrating the advantages of ISAC- Fi’s monostatic sensing according to various example embodiments over conventional bistatic sensing.

[00104] In an experiment performed, the robot car was moved to 16 preset locations in a 100m ² lab space. Both ISAC-Fi and mD-Track operate 3 antennas in 2.4 GHz with 20 MHz bandwidth, capturing a 3 x 3 x 64 CSI matrix from each packet preamble. Subsequently, channel features were jointly estimated to derive AoA, ToF, and hence the location. For each location, 40 measurements were averaged to derive an error by comparing with the ground truth, and 100 errors were collected with 4,000 measurements. As ISAC-Fi excels at ToF estimation, the CDFs of both localization and ToF errors are reported in FIGs. 22A and 22B, comparing mD-Tracks with two ISAC-Fi localization schemes. In this regard, while ISAC-Fii leverages both ToF and AoA estimated by a single device to infer location, ISAC-Fi2 exploits two collaborative devices to reach the same goal. In particular, FIGs. 22A and 22B show the performance of non-collaborative (ISAC-Fii) and collaborative (ISAC-Fi2) MIMO localization. As can be seen, ISAC-Fi outperforms mD-Track with a median localization error down to 1.12 m as opposed to mD-Track’s 4.57 m, and ISAC-Fi2 performs slightly better than ISAC-Fii due to the collaborative sensing. These results are consistent with the analysis described hereinbefore under sub-heading “Uncertainties in Temporal Features’" that ToFs of propagation paths cannot be accurately estimated under the bistatic mode and that ISAC-Fi is designed to address this technical challenge.

Human Activity Recognition

[00105] Contact-free human activity recognition (HAR) plays a key role in a wide range of real-world applications, and existing Wi-Fi-based HAR solutions directly translate CSI to classification results. However, due to the ambiguities of bistatic motion sensing mentioned hereinbefore under sub-heading “Ambiguity in Motion Sensing", such translations can be misled and thus resulting in degraded performance. Therefore, various example embodiments seek to demonstrate using experiments that ISAC-Fi’s monostatic sensing, albeit relying on only a single Wi-Fi device, may achieve comparable or even better performance than existing bistatic solutions with at least two devices involved.

[00106] In fact, basic HAR may not be a perfect task for evaluating sensing capabilities, because conventional bistatic Wi-Fi systems may still yield a high accuracy by overfitting the training/validation data. Therefore, a more difficult cross-domain HAR task was chosen for evaluation, where cross-domain means that the environments and human subjects used in training and testing can be different. It is known that Wi-Fi signals carry a substantial amount of environment and subject specific information, so a Wi-Fi HAR method has to resolve this entangled information in order to generalize to new domains. Consequently, the environment independent (El) framework (disclosed in Jiang et al.. “Towards Environment Independent Device Free Human Activity Recognition”, In Proceedings of the 24 ^th ACM MobiCom 2018, pages 289-304) was selected as the comparison baseline given (i) its cross-domain capabilities achieved by the novel adversarial learning, and (ii) its minimal ITx - 2Rx multistatic setup for Wi-Fi sensing.

[00107] Experiments were conducted under several typical indoor settings. Both ISAC-Fi and the El framework sent 40 packets per second for 10 seconds, and 64-subcarrier CSIs were extracted. We collect a cross-domain dataset by letting 6 male and 4 female subjects perform 6 activities in 10 rooms with different layouts and sizes (ranging from 6 to 50m ² ). Without loss of generality, the activities include sitting down, standing up, walking, falling down, bending, and lying down. Each activity was performed 2,500 times, and a total of 15,000 examples of these activity classes were obtained. The same classifier architecture adopted by the El framework was employed, which includes a 3 -layer convolutional network, a domain discriminator, and respective losses to achieve environment and subject independence.

[00108] FIGs. 23 A and 23B show the confusion matrices of HAR. The evaluation results shown in FIGs. 23 A and 23B indicate that the average HAR accuracy of ISAC-Fi is above 82%, while that of the El framework is less than 72%. The inferior performance of the El framework can be largely explained by its incompetent cross-domain classification ability, which in turn results from the errors brought by the motion sensing ambiguities typical for a bistatic architecture. The results clearly highlight the efficacy of ISAC-Fi according to various example embodiments in resolving such ambiguities, allowing it to achieve a higher accuracy in crossdomain HAR. Wi-Fi Imaging

[00109] Wi-Fi imaging uses Wi-Fi signals for reconstructing images of subjects, which has attracted an increasing attention in recent years. Various existing proposals aim for generating subject images by leveraging various techniques such as large-scale MIMO, synthetic aperture radar, and multistatic setup. Since these proposals often rely on increasing antenna numbers to improve performance, it is almost impossible to quantitatively compare among them. Consequently, only the feasibility of imaging with ISAC-Fi’s monostatic sensing according to various example embodiments is demonstrated below. Specifically, the deep learning techniques adopted by Wang et al., “Person-in-WiFi: Fine-grained Person Perception using WiFi”, In Proceedings of the 33rd IEEE ICCV, 2019, pages 5452-5461 were employed to translate CSIs captured by ISAC-Fi towards images outlines and skeletons of the subjects.

[00110] Experiments were conducted in rooms and corridors. To enable Wi-Fi imaging, ISAC-Fi leverages its 3-antenna array to improve the spatial diversity in perceiving a subject. Human subjects were requested to pose differently at various distances and angles. For each scene, ISAC-Fi averages over 300 packets to obtain a 3 * 3 x 150 CSI matrix, where the first two 3’s refer to the antenna number and 150 indicates that 3 packets as a group with only 50 out of 64 subcarriers per packet are used. Meanwhile, a camera next to ISAC-Fi captures a ground truth photo. Since outlines and skeletons of human subjects are of interest, the photos were further processed to generate binary masks and skeletons of the human subjects for training purposes. A total of 10,000 CSI-image pairs were collected for training deep neural network.

[00111] Since all spatial information was embedded in the CSI samples, it is viable to reconstruct human image (outline) and skeleton from its corresponding CSI sample leveraging the deep learning network designed in the above-mentioned Wang reference. The network treats the 3x3x 150 CSI samples as 3x3 images with 150 channels. It trains a U-Net and skeleton association algorithm to map the CSIs to outlines and skeletons, leveraging the training data created from ground truth photos. For the training process, the batch size was set to 32 and the Adam optimizer was used, whose learning rate and momentum were set to 0.001 and 0.9, respectively. FIG. 24 shows the imaging results of human subjects. In particular, FIG. 24 shows the ground truth photos, RF outlines, and RF skeletons of one, two, and three subjects, respectively. The RF images correctly indicate the number of subjects and clearly show the torso, head, and limbs of each subject, while the skeleton images provide an even sharper characterization of the joint and limb positions. All these results confirm the imaging capabilities of the monostatic sensing adopted by ISAC-Fi according to various example embodiments of the present invention. In various example embodiments, more realistic imaging results can be achieved by combining both monostatic and bistatic sensing.

RELATEDWORK AND DISCUSSIONS

[00112] As explained hereinbefore, Wi-Fi sensing leveraging CSI can be categorized into device-based and device-free methods. Whereas device-based methods are only applicable to locating other user-held Wi-Fi devices, device-free methods impose no requirement on users (e.g., does not require users to hold a Wi-Fi device) but entail a bistatic or even multistatic setting (i.e., involving at least two devices). Moreover, existing proposals on device-free Wi-Fi sensing have largely remained as experimental prototypes because sensing algorithms are often at odd with Wi-Fi communications. In contrast, the ISAC-Fi according to various example embodiments is advantageously proposed and implemented to address these technical challenges faced by existing device-free methods, and it also seeks to seamlessly integrate sensing with communication (normal Wi-Fi communication) so as to realize an ISAC-ready Wi-Fi system or device, including a prototype thereof.

[00113] Furthermore, ISAC-Fi is fundamentally different from full-duplex radios (FDR) as explained hereinbefore under sub-heading “Tx-Rx Signal Separator Design" as it separates Tx (communication) signals from Rx (sensing) signals. As an example, compared with Hassani et al., "In-Band Full-Duplex Radar-Communication System," in IEEE Systems Journal, vol. 15, no. 1, pp. 1086-1097, March 2021, Hassani et al. merely migrates FDR technique to ISAC scenarios without paying attention to their fundamental differences. In addition, Hassani et al. did not consider the compatibility with Wi-Fi framework and it relies on a proprietary chip for Tx-Rx signal separation, which strongly confined its practical feasibility. On the contrary, the ISAC-Fi prototypes according to various example embodiments are clearly implementable as extended Wi-Fi NICs, especially with support from manufacturers. Various example embodiments also seek to realize a large-scale MIMO front-end, obtain a much wider operational bandwidth, and achieve widescale distributed MIMO.

[00114] Accordingly, various example embodiments advantageously provide Wi-Fi ISAC- ready as well as providing two prototypes of ISAC-Fi as implementation examples. Therefore, a more practical paradigm for Wi-Fi sensing is advantageously achieved according to various example embodiments of the present invention. [00115] While embodiments of the invention have been particularly shown and described with reference to specific embodiments, it should be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the scope of the invention as defined by the appended claims. The scope of the invention is thus indicated by the appended claims and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced.