Real-time processing method for a sound signal and sound signal capture device

Abstract

TITLE: Method for real-time processing of a sound signal and sound signal capture device implementing such a method. The present invention relates to a method for processing a sound signal, comprising capturing a sound signal (110), which may contain human voices, applying in real time to said sound signal (110) a transformation isomorphic to the "absolute value" transformation to obtain a processed signal (120), and analyzing the processed signal (120). The invention also relates to a sound signal capture device (100) applying to the sound signal (110) a transformation isomorphic to the "absolute value" transformation. Abstract figure: Figure 1

Description

Title of the invention: Method for real-time processing of a sound signal and device for capturing a sound signal. Field of the invention

[0001] The present invention relates to the capture and processing of a sound signal, for the purpose of its analysis.

[0002] In particular, the invention relates to the processing methods which are applied, in real time, to the signal captured by a sensor such as a microphone, before the steps of analyzing this signal.

[0003] The invention also relates to devices for capturing a sound signal, comprising at least one sensor and equipment for processing the sound signal measured by these sensors. Prior art

[0004] There is a wide variety of sound signal capture devices, which serve different purposes.

[0005] Some of these sound signal capture devices are integrated into devices designed to record, amplify, or remotely transmit a sound signal audible to the human ear, representative of the sound signal captured by a microphone integrated into the sound signal capture device. In such devices, which are implemented, for example, in tape recorders, telephones, audio intercoms, etc., the processing performed on the signal captured by the microphones generally aims to make the sound signal more easily audible. Such processing may, for example, include frequency filters that favor the sound frequencies corresponding to the sounds that one wishes to transmit and / or record, in order to improve the signal-to-noise ratio of the sound signal.

[0006] The present invention does not apply to such sound signal capture devices intended to transmit and / or record an audible sound signal.

[0007] Sound signal capture devices are also known, which are integrated into devices designed to analyze the captured sound signal in order to recognize language elements for decoding an audible sound message, or to identify a voice. Such speech recognition generally requires the speaker's prior authorization.

[0008] The present invention also does not apply to such sound signal capture devices, often called speech recognition devices.

[0009] We also know of some devices for capturing a sound signal whose objective is not to transmit or record an audible sound signal but is, on the contrary, to analyze the physical characteristics of the sound signal on the basis of criteria different from the audible messages that may be contained in this sound signal.

[0010] For example, sound signal capture devices are known for operating outside the frequencies audible to the human ear, such as sonar, which are specifically adapted to capture sounds propagating in a liquid in order to detect and locate objects in that liquid. Similarly, sound signal capture equipment integrated into ultrasound scanners is also known. These devices aim to capture sound signals, generally in the ultrasound range, that are reflected by a propagation medium after being emitted by the ultrasound scanner. Analysis of the reflected sound signal makes it possible to form an image of the medium through which the sound signal propagated.

[0011] The present invention does not apply to such sound signal capture devices, which relate to sound signals that do not include human speech.

[0012] Sound level meters are still known; these are devices for capturing a sound signal, comprising a sensor such as a microphone combined with a processing device designed to determine a normalized sound level, expressed in decibels, of the captured sound signal. In this case, real-time sound signal processing devices may include frequency filters designed to favor frequencies audible to the human ear, so that the measured sound level is representative of the perceived sound level by the human ear. However, such sound signal processing does not allow for analysis of the processed signal that would extract information other than the sound level.

[0013] The present invention does not apply to such sound signal capture devices which only allow superficial analyses of the sound signal.

[0014] Devices for capturing a sound signal aimed at identifying a particular event or anomaly from the captured sound signal are also known. In particular, sound signal capture devices are known that allow the capture of ambient sound in a public space and the identification, from the captured sound signal, of a particular event or anomaly.

[0015] Sound signal capture devices are known that capture ambient sounds in a public space and make it possible to identify, within these ambient sounds, sound events such as explosions, impacts, shouts, etc. Such equipment can, for example, be associated with video surveillance cameras, to specifically alert a video surveillance operator to images captured by the video surveillance cameras covering areas in which a sound signal capture device has detected a sound corresponding to a particular event or anomaly.

[0016] The present invention relates to such sound signal capture devices, intended to be placed so as to capture ambient sounds in a public space, and to allow for analysis of the sound signal based on the physical data of this sound signal that do not contain any language elements. It relates more particularly to such sensors designed to identify, from an ambient sound signal, specific events or anomalies.

[0017] To capture ambient sounds and detect specific events or anomalies, such devices can be placed in locations such as public places, where their sound sensors, such as microphones, can capture information such as human voices. The presence of such devices capable of capturing human voices may legitimately raise concerns about the confidentiality of conversations that would be recorded by the microphones. Sound information, such as the content of conversations, or any information allowing the identification of a person from a voice, is generally considered by law to be personal information, which it is prohibited to record or transmit without the authorization of the individuals concerned.It is therefore generally prohibited to use, in public places, sound recording devices capable of recording or transmitting information containing voices, recognizable language elements or similar signals, even if the device is not intended to perform speech recognition or identification of the speaker.

[0018] The use, particularly in the public domain, of such devices enabling the analysis of ambient sounds to identify a particular event or anomaly, is therefore only possible under conditions guaranteeing, in a particularly reliable manner, the absence of recording or transmission of information that could be considered as personal information.

[0019] Objectives of the invention

[0020] The present invention aims to overcome these drawbacks of the prior art.

[0021] In particular, the present invention aims in particular to enable the use of sound signal capture devices to identify a particular event or anomaly, from the captured sound signal, under conditions such that it can be guaranteed that this device does not transmit or record any personal information contained in the sound signal it captures.

[0022] Another objective of the invention is to provide a processing method, which can be implemented in a sound signal capture device, which allows a sound signal captured by a sensor such as a microphone to be processed in order to reliably and irreversibly remove information that can be considered personal information.

[0023] The invention further aims to provide such a method for processing a sound signal which is capable of performing in real time the processing of a sound signal captured by the sensors.

[0024] The invention further aims to provide such a method for processing a sound signal which provides a processed signal, derived from the sound signal, containing sufficient information from the sound signal initially captured and not relating, with certainty, to personal information, to allow the detection and categorization of particular sound events or anomalies.

[0025] Another objective of the invention is to provide a sound signal capture device, which is capable of capturing a sound signal and processing this sound signal in real time, in such a way that the processed signal, provided by this capture device, does not contain information that could be considered as personal information. Description of the invention

[0026] These objectives, as well as others which will become clearer later, are achieved according to the invention using a method for processing a sound signal, comprising: - the capture of a sound signal, which may contain human voices, - the real-time application, on said sound signal, of a transformation isomorphic to the "absolute value" transformation, to obtain a processed signal, - the analysis of said processed signal.

[0027] Such a process applies a treatment for anonymizing the sound signal. This process makes it possible to obtain a processed signal exhibiting a large number of physical characteristics that allow for its analysis, and in particular the preservation of all its frequency components, in which it is nevertheless impossible to recover any information that could be considered personal information, such as language elements, or characteristics that allow for voice recognition. The processed signal is thus "anonymized," that is to say, it no longer contains the personal information that may have been contained in the sound signal.

[0028] According to an advantageous embodiment, this transformation consists of associating each value of the sound signal with its absolute value.

[0029] Such a transformation is very easy to implement, whether on an analog or digital signal, and its real-time implementation requires very few resources. However, it ensures the irreversible loss of information that could be considered personal information.

[0030] According to one possible embodiment, this transformation consists of associating with each value of the sound signal is the square of its value.

[0031] A person skilled in the art can also imagine an infinite number of other transformations isomorphic to the "absolute value" transformation.

[0032] Advantageously, the process includes digitizing the sound signal, and the transformation is applied to the digitized sound signal.

[0033] Preferably, the analysis of the processed signal includes at least one frequency analysis step of the processed signal.

[0034] Such an analysis, which can provide information on the nature of the sound captured by the sensor, is possible on the signal processed according to the present invention, because this signal is neither averaged nor smoothed. It is impossible on a signal that has undergone averaging or integration, which has the effect of eliminating the oscillations of the initial sound signal.

[0035] Advantageously, the analysis of the processed signal is carried out after a recording of this processed signal.

[0036] Such recording of the processed signal is not prohibited by legislation protecting privacy, insofar as this signal does not include information that can be considered personal information.

[0037] The invention also relates to a sound signal capture device, comprising a pressure sensor and a processing equipment, this processing equipment applying to the sound signal from the sensor a transformation isomorphic to the "absolute value" transformation.

[0038] Such a sound signal capture device makes it possible to provide a processed signal which presents a large part of the information of the sound signal, while not including any information which could be considered as personal information.

[0039] Preferably, such a sound signal capture device consists of a single electronic component, comprising a pressure sensor and an electronic chip ensuring the digitization of the signal from the sensor and the application to this signal of the isomorphic transformation to the "absolute value" transformation.

[0040] Such an electronic component, combining a pressure sensor and a chip in a single package that can be implemented in an electronic circuit, is similar to components known as "analog microphones." However, compared to these analog microphones, they are modified so that the chip, in addition to digitizing the sound signal from the sensor, applies an isomorphic transformation to the "absolute value" transformation of this signal.

[0041] List of figures

[0042] The invention will be better understood upon reading the following description of preferred embodiments, given by way of simple figurative and non-limiting example, and accompanied by figures including: - Fig. 1 is a schematic representation of a device for capturing, processing and analyzing sound signals according to one embodiment of the invention. - Fig. 2 is a graphical representation of a sound signal that can be captured by a sensor such as a microphone. - Fig. 3 is a graphical representation of a processed signal, obtained by applying a processing method according to an embodiment of the invention to the sound signal represented by Fig. 2. - Fig. 4 is a graphical representation of a processed signal, obtained by applying a processing method according to another embodiment of the invention to the sound signal represented by Fig. 2. - Fig. 5 is a schematic representation of an electronic circuit that can implement a processing method according to an embodiment of the invention. Detailed description of embodiments of the invention

[0043] Fig. 1 represents schematically a device for capturing, processing and analyzing a sound signal intended to detect and categorize, in that sound signal, a particular event or anomaly.

[0044] This device includes a sensor 11, for example consisting of a microphone, capable of capturing a sound and generating a signal 110, hereafter called a "sound signal", representative of that sound.

[0045] It should be noted that sounds, within the meaning of the present invention, are preferably audible sounds capable of containing a human voice, which are generally between 200 Hz and 3 kHz. Indeed, only sounds containing human voices are generally considered to contain personal information.

[0046] Figure 2 is a graphical representation of an extract of a sound signal that can be captured by a sensor 11 such as a microphone. Typically, the sound signal 110 provided by the sensor is a continuous signal represented by the curve 21 oscillating around a neutral value 20. This curve 21 is representative of the pressure variations, around a neutral value 20, that characterize the sound.

[0047] The sound signal 110 can be an analog signal directly from the sensor 11. It can also, in a conventional way, have undergone initial pre-processing, such as pre-amplification and filtering by a bandpass filter, intended to remove the continuous component of the signal and to correct the frequencies which are poorly processed by the sensor 11. It is also possible that the sound signal has been digitized.

[0048] In the device of [Fig. 1], the sound signal 110 generated by the sensor 11 is sent to a processing equipment 12 which applies in real time to the sound signal 110 a processing aimed at removing from this signal any information that could be considered as personal information.

[0049] To this end, the processing equipment 12 subjects the sound signal 110 to a transformation isomorphic to the "absolute value" transformation. Thus, the processed signal 120, resulting from this processing of the sound signal 110 by the processing equipment 12, makes it possible to recover the absolute value of the sound signal 110 from the sensor 11.

[0050] Such a transformation is an even, non-averaged function. An even function, applied to a variable signal oscillating around an equilibrium value considered as the origin, is a function that associates with each value of the signal, positive with respect to the origin, the same result as for the same value of the signal, negative with respect to the origin. The even function thus generates an image that varies according to the absolute value of the original signal. According to the invention, this function is non-averaged, that is to say, at any given instant, the generated image is a function only of the absolute value of the original signal at that precise instant. This function therefore makes it possible to take into account all the oscillations of the original signal, without smoothing these oscillations.

[0051] The inventors observed that applying the "absolute value transformation," or a transformation isomorphic to the absolute value, irreversibly alters a sound signal. The term "absolute value transformation" refers to a transformation that associates the absolute value of a number with that number. It is thus impossible to recover the original sound signal from the processed signal. Furthermore, the processed signal irreversibly loses its information that allows for the identification of a sound or a tone of voice. The processed signal is therefore irreversibly anonymized and no longer contains any information that could be considered personal data.

[0052] The processed signal 120, resulting from this transformation performed by the processing equipment 12, is then sent to an analysis unit 13. This analysis unit 13 can be used to analyze the processed signal 120, preferably in real time, in order to identify anomalies or specific events in the sound signal captured by the sensor 11. This processed signal 120 has the significant advantage of preserving all the oscillations of the initial sound signal 110, its frequency being doubled compared to the frequency of the initial signal. It is therefore possible to perform a frequency analysis, for example, to categorize the different sound components of the initial sound signal.

[0053] In other embodiments, it is also possible that the processed signal is sent to recording equipment in order to record the processed signal 120, for example to allow its subsequent analysis. It is also possible that the the processed signal is sent to a transmission device, in order to transmit this processed signal 120 to another device, which can for example carry out its analysis in order to detect and categorize particular events or anomalies in the sound signal captured by the sensor 11.

[0054] It should be noted that a transmission or recording of the sound signal 120 should not be limited by legislation restricting the transmission or recording of data containing personal information, since this sound signal 120 does not contain information that could be considered personal information.

[0055] Figure 3 is an example of a representation of a processed signal, resulting from the processing of the sound signal represented by Figure 2, by a processing unit according to an embodiment of the invention. In this embodiment, the processing unit applies an "absolute value" transformation, consisting of associating each value of the sound signal with its absolute value, relative to the equilibrium value. All the values ​​of the processed signal, represented by curve 22, are therefore positive with respect to the neutral value 20.

[0056] Such an "absolute value" transformation can, for example, be obtained on an analog audio signal by subjecting this signal emitted by a sensor such as a microphone to an electronic circuit such as a thresholdless rectifier, well known to those skilled in the art. An example of such an electronic circuit, comprising an operational amplifier 43, a diode 44, and a resistor 45, is shown in [Fig. 4]. Such a circuit makes it possible to associate an input voltage 41 with an output voltage 42 whose value corresponds to the absolute value of the input voltage 41.

[0057] Such an "absolute value" transformation can also be obtained very easily on a digitized sound signal, by applying a program adapted to this signal.

[0058] Figure 5 is an example of a representation of a processed signal, resulting from the processing of the sound signal represented by Figure 2, by another processing unit according to an embodiment of the invention. In this embodiment, the processing unit applies a "square" transformation, consisting of associating each value of the sound signal with the square of its value. All the values ​​of the processed signal, represented by curve 32, are therefore positive with respect to the neutral value 30.

[0059] The "square" transformation is isomorphic to the "absolute value" transformation. It is indeed possible, by calculating the square root of the processed signal, to obtain the absolute value of the initial sound signal.

[0060] Such a “square” transformation can also be obtained very easily on a digitized sound signal, by applying a program adapted to this signal.

[0061] It should be noted that a person skilled in the art could imagine an infinite number of transformations The sound signal isomorphic to the "absolute value" transformations. These transformations will have the same effect of irreversibly removing all information from the sound signal that could be considered personal information. Furthermore, they will also have the advantage of preserving the oscillations of the sound signal and allowing for its frequency analysis. Such frequency analysis advantageously allows for greater precision in the detection and / or categorization of an event or anomaly in an ambient sound signal.

[0062] A sound signal capture device according to the invention may combine a pressure sensor such as a microphone and processing equipment applying to the signal from this sensor a transformation isomorphic to the "absolute value" transformation. Such a sound signal capture device is schematically represented in [Fig. 1] by rectangle 100.

[0063] According to a particularly advantageous embodiment, this sound signal capture device may consist of a single electronic component, comprising a pressure sensor and an electronic chip that digitizes the signal from the sensor and applies to this signal a transformation isomorphic to the "absolute value" transformation. This electronic component may also include means for preprocessing the signal, such as preamplification and filtering by a bandpass filter, for example before its digitization.

[0064] Such an electronic component may advantageously be similar to a component known as an "analog microphone," which comprises a pressure sensor and an electronic chip that digitizes the signal from the sensor. This sensor and chip can be housed in the same package as the electronic component, which can itself be implemented in an electronic circuit. To conform to the invention, the chip is programmed in this case to also perform a transformation isomorphic to the "absolute value" transformation. Such an electronic component can thus capture a signal, representative of ambient sound, which can undergo frequency analysis, while ensuring that this signal does not contain any information that could be considered personal information.

[0065] Furthermore, combining the sensor and the signal processing chip in the same housing advantageously ensures that the unprocessed audio signal is not used. Implementing such a microphone in an electronic circuit thus guarantees that this electronic circuit does not use any audio data that could be considered personal information.

Claims

Claims

1. Method for processing a sound signal, comprising: - capturing a sound signal (110), likely to contain human voices, - applying in real time, to said sound signal, a transformation isomorphic to the “absolute value” transformation, to obtain a processed signal (120), - analyzing said processed signal (120), characterized in that said analysis of said processed signal (120) comprises at least one step of frequency analysis of said processed signal (120).

2. Processing method according to the preceding claim, characterized in that said transformation consists of associating with each value of the sound signal (110) its absolute value.

3. Processing method according to claim 1, characterized in that said transformation consists of associating with each value of the sound signal (110) the square of its value.

4. Processing method according to any one of claims 1 to 3, characterized in that said method comprises digitizing said sound signal (110), and in that said transformation is applied to said digitized sound signal.

5. Processing method according to any one of claims 1 to 4, characterized in that said analysis of said processed signal (120) is carried out after recording of said processed signal (120).

Images