WIRELESS COMMUNICATION SYSTEMS

TECHNICAL FIELD

[0001] In some example embodiments, the subject matter herein generally relates to wireless communication systems, and more specifically to an efficient learning algorithm for channel estimation in wireless communication systems.

BACKGROUND

[0002] Modem day wireless communication systems conforming to standards like 5G New Radio (NR) transmit and receive signals in units of time called slots. Each transmitted slot of signal travels through a wireless propagation channel before reaching the receiver. The effects of channel depend on environmental topography as well as the relative velocity of the receiver with respect to trans itter.

[0003] Each slot of signal received at the receiver is independently equalized to compensate for the effects of channel. As the channel is estimated independently on each slot, traditional signal processing algorithms fail to utilize the information that the receiver already has about the channel in past slots. That makes the estimated channel noisy and the solution thus built does not perform to the full potential, resulting in drop in throughput in the low signal to noise (SNR) region.

[0004] The primary reason for avoiding the usage of knowledge of the channel over past slots is the possibility of time selective or time varying channels. Depending on the relative velocity and the consequent Doppler spread the channel behavior can vary with time and it becomes extremely difficult to gauge the extent to which information of the channel over past slots can be reliably used as an input to estimate the channel in a given slot. Thus, current channel estimation methods generally do not make use of the knowledge of the channel over past slots. [0005] However, even when the channel is time selective, it is mainly the phase of the complex channel coefficients that vary across slots. The power delay profile would remain unchanged at least over a few tens of milliseconds. So irrespective of the channel conditions there is useful information in the channel over past slots that can be used to improve the channel estimate in a given slot. Thus, there is a need for an enhancement of channel estimation by making use of receiver’s knowledge of channel in past slots without increasing the real-time processing requirements.

SUMMARY

[0006] In one general aspect, a method of estimating a wireless channel is provided. The method includes determining a value of [3 from a finite set of discrete values ranging from 0 to 1 to be used for channel estimation in a slot number N based on time-selectivity of a channel corresponding to the slot number N; estimating the channel according to reference signal symbols received in the slot number N; determining a transmit power level in a slot number N - 1 ; determine a transmit power level in the slot number N; determining a power offset ratio of the transmit power in the slot number N - 1 to the transmit power in the slot number N; identifying filtered channel coefficients in the slot number N - 1 ; and computing in real time filtered channel coefficients in the slot number N according to a determined set of values [3 applied to generate a subset of equalized QAM symbols in the slot number N, the estimated channel according to reference signal symbols received in the slot number N, the determined power offset ratio, and the filtered channel coefficients in the slot number N - 1 .

[0007] Implementations of the method may include one or more of the following features. In the method, a value of 1 indicates reliance on a previous channel estimation and a value of 0 indicates reliance on current channel estimation, according to an embodiment. Further in the method, computing in real time filtered channel coefficients in the slot number N includes equalizing a subset of complex QAM symbols with each value in the set of values for f3, according to an embodiment.

[0008] In another general aspect, a wireless channel estimating apparatus is provided. The wireless channel estimating apparatus includes a memory configured to store a plurality of instructions; processor circuitry coupled to the memory and configured to execute the plurality of instructions to determine a value of [3 from a finite set of discrete values ranging from 0 to 1 to be used for channel estimation in a slot number N based on time-selectivity of a channel corresponding to the slot number N; estimate the channel according to reference signal symbols received in the slot number N; determine a transmit power level in a slot number N - 1 ; determine a transmit power level in the slot number N; determine a power offset ratio according to the ratio of transmit power level in the slot number N - 1 to the transmit power of the slot number N; identify filtered channel coefficients in the slot number N - 1 ; and compute in real time filtered channel coefficients in the slot number N according to a determined set of values [3 applied to generate a subset of equalized QAM symbols in the slot number N, the estimated channel according to reference signal symbols received in the slot number N, the determined power offset, and the filtered coefficients in the slot number N - 1 .

[0009] Implementations of the wireless channel estimating apparatus may include one or more of the following features. In the wireless channel estimating apparatus, [3 is a momentum term and depends on the time selectivity of the channel, and wherein an optimum value of [3 depends of the time selectivity of the channel, according to an embodiment. In a further embodiment, a value of 1 indicates reliance on a previous channel estimation and a value of 0 indicates reliance on current channel estimation. In a further embodiment, computing filtered channel coefficients in real time in the slot number N includes equalizing a subset of complex QAM symbols with each value in the set of values for f3, and in a further embodiment the value of [3 that produces the highest mean value of absolute Log-Likelihood Ratio (LLR) of demodulated QAM symbols is chosen as the best value of [3 for slot number N. In a further embodiment, the real time filtered channel coefficients in the slot number N are computed using the best value of [3 from the set of values [3 for equalization of all the complex QAM symbols allocated. In a further embodiment, a subset of complex QAM symbols is randomly chosen over an entire bandwidth. In a further embodiment, a cardinality of the subset of QAM symbols is much smaller than a total number of QAM symbols allocated. In a further embodiment, the processor circuitry iteratively equalizes all the complex QAM symbols estimated using each value of [3 from the finite set of discrete values in attempting to decode a payload for every iteration. [0010] In another general aspect, a wireless channel estimating system is provided. The wireless channel estimating system includes a transmitter configured to transmit at least one reference signal and a plurality of Quadrature Amplitude Modulated (QAM) data symbols; a receiver configured to receive the at least one reference signal and the plurality of QAM data symbols; determine a value of [3 from a finite set of discrete values ranging from 0 to 1 to be used for channel estimation in a slot number N based on time-selectivity of a channel corresponding to the slot number N; estimate the channel according to the at least one reference signal symbols received the slot number N; determine a transmit power level in a slot number N - 1 ; determine a transmit power level in the slot number N; determine a power offset ratio between the slot number N - 1 and slot number N; identify filtered channel coefficients in the slot number N - 1 ; and compute in real time filtered channel coefficients for the slot number N.

[0011] Implementations of the wireless channel estimating system may include one or more of the following features. In the wireless channel estimating system, the real time filtered channel coefficients are computed according to the following equation according to an embodiment: where N is the slot number; [3 is the momentum of a propagation channel comprising the slot N; H _iltered is a vector of the filtered channel estimate in the slot N; Hg _Stimated is a channel estimated using only the QAM symbols of the at least one reference signal received in slot N; is a vector of the filtered channel estimate in slot N - 1 ; and Power Offset is the ratio of the transmit power in slot N - 1 to the transmit power in slot number N. In a further embodiment, a set of values for [3 are determined according to the time selectivity of the channel, and wherein a value of 1 indicates reliance on a previous channel estimation and a value of 0 indicates reliance on a current channel estimation. In a further embodiment, computing filtered channel coefficients in real time in the slot number N includes equalizing a subset of complex QAM symbols with each value in the set of values for [3. In a further embodiment, the value of [3 that produces the highest mean value of absolute Log-Likelihood Ratio (LLR) of demodulated QAM symbols is chosen as the best value of [3 for slot N. In a further embodiment, the real time filtered channel coefficients in the slot number N are computed using the best value of [3 from the set of values [3 for equalization of all the complex QAM symbols allocated. In a further embodiment, a subset of complex QAM symbols is randomly chosen over an entire bandwidth, and wherein a cardinality of the subset of QAM symbols is much smaller than a total number of QAM symbols allocated. In a further embodiment, the receiver is configured to iteratively equalize all the complex QAM symbols estimated using each value of [3 from the finite set of discrete values in attempting to decode a payload for every iteration.

BRIEF DESCRIPTION OF THE DRAWINGS

[0012] FIG. 1 illustrates a process of estimating a wireless channel according to an embodiment.

DETAILED DESCRIPTION

[0013] In channel estimation, one of the most promising ideas is to use neural networks and allow them to adapt the weights for past and the present inputs dynamically. There is a lot of study that is ongoing in this regard and excellent improvements in performance have already been reported and published (see, for example, Refs. [1 ] - [3]). However, computing weights of a neural network in every slot is an extremely costly operation in the current generation of processors that host baseband signal processing operations.

[0014] An embodiment of the present invention proposes an efficient and yet low- cost alternative as described below.

[0015] FIG. 1 shows a process flow for estimating a wireless channel according to an embodiment.

[0016] In 110, a value [3 is determined from a finite set of discrete values (0 < [3 < 1 ) that is to be used for channel estimation in a slot number N based on time-selectivity of a channel corresponding to the slot number N. Here, [3 is known as the momentum of the propagation channel.

[0017] In 120, the channel is estimated according to reference signal symbols received in the slot number N. This gives the current Hg _Stimated , which is a channel instantaneously estimated using only the QAM symbols of the at least one reference signal received in slot N.

[0018] In 130, a transmit power level in a slot number N - 1 is determined, and in 140 a transmit power level in the slot number N is determined. Using the determined power levels in 130 and 140, a power offset ratio between the slot number N - 1 and slot number N is determined in 150.

[0019] Useful information in the channel over past slot N - 1 can be used to improve the channel estimate in the present slot N. In 160, filtered channel coefficients in the slot number N - 1 are identified. Here, H^ _t g _red is a vector of the filtered channel estimate in slot N - 1 .

[0020] In 170, filtered channel coefficients in the slot number N are computed in real time. Here, H^ _iltered is a vector of the filtered channel estimate in the slot N, which is computed in real time based on a determined set of values [3 applied to generate a subset of equalized QAM symbols in the slot number N, the estimated channel according to reference signal symbols received in the slot number N, the determined power offset, and the filtered coefficients in the slot number N - 1 .

[0021] In one embodiment, the real time filtered channel coefficients are computed according to the following equation: where N is the slot number; [3 is the momentum of a propagation channel comprising the slot N; Hf _iltered is a vector of the filtered channel estimate in the slot N; Hg _Stimated is a channel estimated using only the QAM symbols of the at least one reference signal received in slot N; H^ _red is a vector of the filtered channel estimate in slot N - 1 ; and Power Offset is the ratio of the transmit power in slot N - 1 to the transmit power in slot number N.

[0022] As can be seen from the right hand side of Eq. 1 , when f3 = 1 , the second term goes to zero, indicating that the filtered channel estimate in the slot N relies solely on the filtered channel estimate in the slot N - 1 ; whereas when [3 = 0, the first term goes to zero, indicating that the filtered channel estimate in the slot N relies solely on the instantaneous channel estimate in the slot N. [0023] Note that the right value for [3 depends on the time selectivity of the channel, and Hf _iltered is estimated for a set of values of f3, according to an embodiment. In one embodiment, a subset of complex QAM symbols is equalized using channel estimated with each value of f3, according to an embodiment. According to an embodiment, the value of p that gives the highest mean of absolute Log-Likelihood Ratio (LLR) values post demodulation of QAM symbols is chosen as the best value for the given slot. In one embodiment, the best value of p for a specific resource block is also identified to be the best value for the entire allocated bandwidth for transmission in a slot as p depends only on the relative velocity. This property of p enables the subset of complex QAM symbols used for p determination to be chosen randomly over the entire allocated bandwidth. In one embodiment, cardinality of this subset of QAM symbols can be chosen to be much smaller than the total number of QAM symbols in the allocation and depends on the allocation size as well as affordable million instructions per second (MIPS) for p determination.

[0024] Depending upon the available physical cores in the system, a superior solution is to iteratively equalize all the complex QAM symbols with channel estimated using every value in the finite set of discrete values for p and attempt to decode the payload in every iteration, according to an embodiment.

[0025] It is observed in simulations that an embodiment of the present invention gives significant increase in the performance of the receiver when the received signal is weak as it is the case typically for Cell edge users. Note that efficiency of the proposed channel estimation algorithms and the associated performance are close to that of neural networks-based learning algorithms. An embodiment also does so without significantly increasing the MIPS requirement of the traditional channel estimation algorithms. Thus, embodiments of the present invention provide unique advantages over existing methods and systems.

[0026] Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed features, from a study of the drawings, the disclosure, and the appended claims.

[0027] In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article “a” or “an” does not exclude a plurality. [0028] A single processor, device or other unit may fulfill the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

[0029] Operations like acquiring, accessing, analyzing, capturing, comparing, determining, displaying, inputting, obtaining, outputting, providing, store or storing, calculating, simulating, receiving, warning, and stopping can be implemented as program code means of a computer program and/or as dedicated hardware.

[0030] A computer program may be stored and/or distributed on a suitable medium, such as an optical storage medium or a solid-state medium, supplied together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunication systems.

REFERENCES

[0031]

Ref. [1]: H. Ye, G. Y. Li and B.-H. Juang, “Power of deep learning for channel estimation and signal detection in OFDM systems,” IEEE Wireless Communications Letters (Volume: 7, Issue: 1 , Feb. 2018)

Ref. [2]: Xuemei Yi and Caijun Zhong, “Deep Learning for Joint Channel Estimation and Signal Detection in OFDM Systems”, IEEE Communications Letters (Volume: 24, Issue: 12, Dec. 2020)

Ref. [3]: David Neumann; Thomas Wiese; Wolfgang Utschick, “Learning the MMSE Channel Estimator”, IEEE Transactions on Signal Processing (Volume: 66, Issue: 11 , Junel , 1 2018)