Login| Sign Up| Help| Contact|

Patent Searching and Data

Document Type and Number:
WIPO Patent Application WO/2019/162824
Kind Code:
A method and a selective noise cancelling system is presented, which is particularly optimized for applications in industrial production plants comprising machinery controlled by blue collars.

Application Number:
Publication Date:
August 29, 2019
Filing Date:
February 19, 2019
Export Citation:
Click for automatic bibliography generation   Help
International Classes:
Foreign References:
Other References:
SOUNDBUBBLE OFFICIAL: "SoundBubble: The tailored solution for industrial environments", YOUTUBE, 20 February 2018 (2018-02-20), pages 1 pp., XP054978851, Retrieved from the Internet [retrieved on 20181112]
Attorney, Agent or Firm:
MOLA, Edoardo et al. (IT)
Download PDF:

1. Digital audio processing method comprising the steps of:

- selecting one or more sounds to be filtered including one or more sounds having, as a relative audio source, a moving mechanical part of a machine (1) and warning sounds generating via an electronic control unit of the machine (1);

- acquiring, in a plant hosting the machine (1), an audio signal comprising one or more sounds via a microphone unit (3) for no more than 3 seconds;

- processing the audio signal so as to:

Separate audio tracks from the relative audio source via a Source separation algorithm; recognize the tracks via a Source Recognition algorithm based on stored identification data about the selected sounds; play via an ear-wearable speaker (2) (in-ear or on-ear o around-the ear), a filtering or canceling audio signal generated on the basis of the steps of separating and recognizing to interfere with the sounds of the step of selecting; and play via the wearable speaker (2) environmental alarm sounds not generated by an electronic control unit of the machine (1) and/ or alarm sounds generated by the control unit of the machine, such sounds having respective pre-stored identification data and not being selectable in the step of selecting.

2. Method according to claim 1 comprising the step of providing the microphone unit (3) as a directional unit placed in a fixed position to acquire sounds coming from an identified machine and coupling in data exchange the microphone unit (3) with the wearable speaker.

3. Method according to claim 2, comprising the step of automatically activating the coupling when a threshold value is reached of a parameter indicating the proximity between the wearable speaker (2) and the microphone unit (3).

4. Method according to claim 1, comprising the steps of providing the microphone unit onboard of the wearable speaker (2) and wherein the microphone unit comprises at least two directional microphones having incident audio axes; of providing a gyroscope onboard of the wearable speaker; and of selectively activating and deactivating the directional microphones based on a signal from the gyroscope to intercept the sounds of the machine (1).

5. Method according to any of the preceding claims, wherein the step of recognizing comprises the step of recognizing one or more sounds generated by a second speaker (3) controlled by the control unit and placed onboard of the machine (1).

6. Method according to any of the preceding claims, further comprising the following step before the step of processing:

- providing a data library to identify said one or more sounds generated by the motion of moving mechanical parts in a machine (1); wherein the sounds in the step of selecting are organized in a list based on the library; and wherein both the separation and the recognition in the step of processing are based on data coded in the library.

7. System comprising:

a user interface to select one or more sounds to be filtered including one or more sounds having, as a relative audio source, a moving mechanical part of a machine (1) and warning sounds generating via an electronic control unit of the machine (1); a microphone unit for the acquisition in a plant hosting the machine (1), an audio signal comprising one or more sounds;

an ear-wearable speaker (2) (in-ear or on-ear o around-the-ear) and a control unit programmed to acquire an audio signal via the microphone unit for no more than 3 seconds and to process the signal so as to:

separate audio tracks from the relative audio source via a Blind Source Separation;

recognize the tracks via a Source Recognition algorithm based on stored identification data about the sounds selected via the user interface;

control the reproduction via an ear-wearable speaker (2) of a filtering or canceling audio signal generated on the basis of the steps of separating and recognizing to interfere with the sounds selected via the user interface; and the reproduction via the wearable speaker (2) of sounds generated by the control unit of the machine, such sounds having respective pre-stored identification data and not being selectable via the user interface.

Audio signal digital processing method and system thereof for use in

a production plant with machinery



The present invention relates to a processing method of the audio signal and to a system thereof, in particular for use in a work environment such as a mechanical workshop or the like for use by blue collars exposed to noises from machine tools or other workshop machinery.


It is known in production plants as workshops with machinery and / or machine tools, to provide passive noise cancellation devices such as ear plugs, useful for dampening noise generated by the movement of mechanical parts of machinery such as bearings, cutting edges, presses. This, however, also dampens sounds that are important for productivity, such as the machine start and end beeps, and sounds important for safety, such as fire alarms etc.

At the same time, it is known to provide active noise cancellation devices to reduce unwanted audio (noise) by generating a second audio specifically designed to interfere and cancel unwanted audio. In particular, this is known in the tourism sector, e.g. headphones to promote rest during flights, and entertainment such as to improve listening to music or to encourage concentration while reading.


The object of the present invention is to process the audio of a production plant with machinery, such as a mechanical workshop, to cancel unwanted audio and increase the concentration of the operator assigned to its own automated work station. The object of the present invention is achieved by a method according to claim 1 and a system according to claim 7.

In particular, the operator selects at will the most annoying sounds of the machinery and the wearable speaker cancels or reduces the effect through active acoustic interference. Moreover, given the substantially simple nature of the selectable sounds, the computing power required for processing the active acoustic interference signal can be obtained by means of a processor placed on board the wearable speaker, which can be an on-ear or around-the-ear headphone and also provide inserts of acoustically insulating material for passive sound damping. For safety reasons, a list of environmental sound alarms is also provided, for example evacuation, fire prevention, leakage of dangerous substances and / or alarms generated by an electronic control unit of the machinery, in which the alarms are different from the acoustic warnings. The identification data of the alarm sounds do not appear in the list and are always reproduced if they are identified in the phases of separating and recognizing.

A configuration is also provided in which the wearable speaker is coupled in data exchange by means of an identification code with a directional microphone having a fixed position such as to pick up the sounds of a single machine. In this way, if there are more identical machines placed side by side and assigned to different operators, the wearable speaker carried by each operator will receive exclusively pertinent signals from their own machinery, reducing the risk of confusion.

Other advantages and features of the present invention are discussed in the description and mentioned in the dependent claims.


The invention is described below on the basis of non-limiting examples illustrated for explanatory purposes in the annexed drawings, in which:

- Figures 1 to 3 refer to respective schematic embodiments of the present invention; and

- Figure 4 refers to user interface screens in which, on the right, different sound filtering modes are shown, including complete silence (except for safety alarms), background noise, warning signals, machinery and, on the left, different types of machines with related work sounds, e.g. a generator, a turbine, a pump, a cutter, a lathe, a press.


Figure 1 schematically shows a work machine 1, for example a machine tool, and a selective noise canceling device 2 worn by an operator of the machine tool or other blue collar.

The machine tool 1 generates a first group of sounds linked to the main operation for which it is designed and generated by the movement of mechanical elements, for example in the case of a lathe all the operations related to the removal of the cutting tools such as the movements for the assembly of the tool, if executed automatically through a tool changer present in the work station, the actual cutting of a piece starting e.g. from a cylindrical semi-finished piece. In particular, it is possible to classify all the sounds emitted by the machine in the various and simpler phases of operation and, in particular, of cutting. For example, loading and unloading of each tool from the automatic tool magazine; cutting by each tool; opening / closing of the safety screen and protection against the cooling fluid; for each tool, cutting of one or more materials for which the tool is compatible; sounds generated by moving fans; sounds generated by bearings etc. A second class of sounds is generated by a loudspeaker 3 present on the machine tool and controlled by an electronic control unit of the work station or of the machine tool 1 to emit acoustic warning signals such as the start of the work cycle, e.g. when the cutting tool operates; end of work cycle, e.g. when the cutting tool finishes cutting and is waiting for a new operation or is placed in the automatic changer.

According to the present invention, an audio recognition system comprising the headset 2, a microphone (not shown) for acquisition of the sounds in the workshop in particular when the machine tool 1 is in operation and a computer that processes the audio signal coming from the microphone and is programmed to perform the following operations:

• Record an audio signal comprising one or more ambient sounds, in particular while the machine tool 1 is in operation;

• Segment or separate the audio signal collected by the microphone to identify a plurality of different sound tracks and process the latter to generate identification data for each track;

• Associate through known similar criteria each track to a predefined number of identification data pre-stored in a system memory and pre-selected by the user as sounds to be deleted or attenuated;

• Generate a sound signal through the headset 2 that interferes destructively with the sound (s) generated by the machine tool 1 that the user has selected as sound (s) to be canceled or attenuated.

It is important that, in the circumscribed environment of the mechanical workshop, the sounds generated by machinery, in particular by machine tools, are characterized by a relatively low number of parameters, such as frequencies and amplitudes. This allows to speed up the separation and classification of the individual sound tracks. Therefore, acquisitions through the microphone having duration of no more than 3 seconds, preferably not more than 2 seconds, allow to reach the desired effect.

According to the embodiment indicated in figure 1, a processor and a memory programmed to perform the above are arranged on board the headphones, which for example are of the 'around the ear' or 'on the ear' type. Likewise, the microphone is also on the headphones 2.

It is also important that the warning signals generated by the machine tool 1, for example through the loudspeaker 3, are not filtered so that the operator assigned to the machine tool 1, and only this operator, listens to these warning signals, for example to avoid confusion if two identical machine tools or two different machine tools but having the same warning signals generate confusion for the relative operators. In particular, a configuration is provided for recognizing and associating a predefined number of sounds with the machine tool generating them.

According to a first embodiment, headset 2 comprises a pair of directional microphones having sound acquisition axes arranged at 90° and a gyroscope for detecting the rotation of the headset 2 with respect to the reference machine tool. The directional microphones are operated on the basis of the angular position signal of the headphone 2 generated on the basis of the gyroscope, and preferably there is an indication for the user about which earpiece must correspond to the right ear and which to correspond to the left ear in a manner that the system associates an angular position of the headset 2 to the face of the operator. One of the two directional microphones has its own acoustic axis parallel to the frontal direction of the operator's face and it is assumed that the operator normally looks at the machine tool 1 during work. If the operator rotates the head, this is detected by the gyroscope. Consequently the second directional microphone is activated and the active microphone up to that moment is deactivated.

According to a further embodiment, shown in Figure 2, a directional microphone 4 is provided mounted in a fixed position, for example on the relative machine tool 1 and coupled in data exchange by an identification code with the headset 2 of the target operator attached to the machine tool 1. In particular, the association in data exchange between the directional microphone 4 and the headset 2 is such that the signals of the loudspeaker 3 and the sounds of the machine tool 1 are detected and sent to the target headset only and not to other headsets 2 worn by colleagues of the operator associated with the machine tool 1. In particular, the directional microphone 4 is oriented to pick up the signals of a single machine tool 1 and avoid picking up signals from adjacent machine tools.

Alternatively, it is possible to associate to the loudspeaker 3 also a remote control signal emitter for the wearable loudspeaker 2, also coupled in data exchange with an identification code to the target headset and configured to send a command signal intercepted only by the target headset for generating, through the wearable loudspeaker 2, a sound perceived by the operator and having the same function as the warning sound signals generated by the loudspeaker 3. In this case, the warning sounds generated by the loudspeaker 3 can be neglected during segmentation or separation and classification or recognition.

According to the embodiment of Figure 3, the workshop comprises a network of antennas N for a wireless exchange of data with a suitable bandwidth, i.e. greater the number of antennas per unit of area of the environment, the greater bandwidth for data exchange. The N antennas are, for example, wi-fi® antennas or Bluetooth® antennas or LoRa antennas. It is important to note that, in order to increase the bandwidth for data exchange, gain, transmission / reception power and calculation power are also involved. In particular in the embodiment of figure 3, the wearable loudspeaker 2 can be of the 'in-ear' type.

In order to save time and reduce the latency of the data process, it is preferable to include a step of providing a library of audio data for learning the point audio sources. This is for example implemented through the acquisition and the process of a sound track relative to each work phase of the machine tool 1 as described above. This data in the library acts as respective dictionaries during separation and recognition.

In addition, the library also includes audio data relating to environmental alarms or alarms generated by the electronic control unit of the machine tool commands and different from the warning sounds.

An example of processing audio tracks in order to extract appropriate dictionary data for the library is the preparation of an audio digital fingerprint by quantization or vector / matrix factoring of the learning audio tracks, in which each dictionary track is processed to generate an identification vector / matrix, ie the learning data, included in the library. During separation and / or recognition, the digital process unit then processes the environmental audio to identify the library identification vector / matrix. It is preferable that the sound tracks of all machine tools in the workshop are processed in order to extract learning data to be added to the library.

For example, the identifying data in the library may comprise, for each sound, a plurality of pairs of audio and reference point fingerprints, in which the latter is identified at a particular instant of time of the audio signal acquired for the library; in which the instant of time is calculated based on the content of the audio signal acquired for the library; and in which the audio digital impression characterizes one or more characteristics, such as frequency spectrum, bass, amplitudes and periodicity, with particular reference to the alarm and warning signals of the acquired audio signal for the library at the reference point;

The step of recognizing or classifying preferably comprises the step of calculating, for each segmented or separate track detected by the microphone, a plurality of pairs of audio digital fingerprint and reference point as described for the identification data in the library; identifying substantially linear correspondences between the plurality of pairs in the library and the plurality of pairs calculated from the audio signal detected by the microphone; and identify library data with a large number of substantially linear matches. For example, this approach is described in greater detail in EP-B1-1307833.

Preferably, the separation or segmentation step is performed by means of a Blind Source Separation (BSS) algorithm, such as a Barra-Spence algorithm. For example, a BSS algorithm is based on multi-channel techniques and Time Direction of Arrival (TDOA) techniques.

The power supply of the headphone 2 can be either battery-powered or by wire.

In all cases where a data exchange coupling with identification code is provided between the headset 2 and an electronic device on board the machine 1, it is possible to provide that at least one of the various sound filtering modes is automatically activated if the headset 2 exceeds a threshold that indicates the adequate proximity of the operator to the machine 1, for example on the basis of the intensify of the signal emitted by the electronic device and picked up by the headset 2. This activation, and optional deactivation when the headset 2 moves too far away from the machine 1, may include a specific combination of sound filtering functions.