Method and device for extracting features from an audio signal

Abstract

A device for extracting features from a sound time-series signal, the device comprising: a capacitor microphone comprising a membrane and a backplate, the microphone for detecting the sound time-series signal and outputting a feature signal indicative of features of data contained in the sound time-series signal; a DC signal generator for generating a DC current signal, a voltage amplitude of the DC current signal being greater than an amplitude of a minimal operational bias voltage of the capacitor microphone; an AC signal generator for generating a periodic AC current signal, a voltage amplitude of the periodic AC current signal being less than the voltage amplitude of the DC current signal; and a mixer for mixing the DC and periodic AC current signals to obtain a mixed current signal and for applying the mixed current signal to the microphone so that the microphone exhibits a limit cycle behavior.

Description

METHOD AND DEVICE FOR EXTRACTING FEATURES FROM ANAUDIO SIGNAECROSS-REFERENCE TO RELATED APPLICATION[0001 j The present application claims priority on US Provisional Patent Application No. 63 / 520,957 filed on August 22, 2023, the content of which is hereby incorporated by reference.FIELD

[0002] The present technology pertains to the field of reservoir computing, and more particularly to methods and systems for extracting features from an audio signal.BACKGROUND

[0003] A major application of physical reservoir computing is sound recognition. However, conventional approaches still require separate electrical or optical setups and data preprocessing. As such, the dissemination and application of physical reservoir computing is hampered by the bulky setup and cumbersome numerical algorithms to handle the data.

[0004] Some conventional physical reservoir computing approaches use time- delayed feedback to conduct time-multiplexing (i.e., combine the previous results processed from physical reservoir with the new data) to feed more information into the reservoir for computation. However, the delayed feedback needs digital to analog conversion, which complicates the physical reservoir setup, adds power consumption, and most importantly, sacrifices the processing speed. Therefore, physical reservoir computing is usually built with optical setups and quantum circuits for higher processing speed which are cumbersome and may be economically unviable for the deployment to edge devices. Additionally, a reservoir usually uses non-linearity that is not associated with the oscillatory nature of audio signals. As such, most of physical reservoir computers require the users to preprocess the audio data using spectrogrambefore feeding to the physical reservoirs. Such approaches usually require processing chips and digital to analog converter which reduces the processing speed and increases the computational power envelope. Therefore, the benefits brought by the analog computing such as high speed and low computational power are reduced or canceled out by the added components to the physical reservoirs.

[0005] Therefore, there is a need for an improved method and system for extracting feature from a sound signal in the context of physical reservoir computing.SUMMARY

[0006] In accordance with a broad aspect, there is provided a device for extracting features from a sound time-series signal, the device comprising: a capacitor microphone comprising a membrane and a backplate, the capacitor microphone for detecting the sound time-series signal and outputting a feature signal indicative of features of data contained in the sound time-series signal; a DC signal generator for generating a DC current signal, a voltage amplitude of the DC current signal being greater than an amplitude of a minimal operational bias voltage of the capacitor microphone; an AC signal generator for generating a periodic AC current signal, a voltage amplitude of the periodic AC current signal being less than the voltage amplitude of the DC current signal; and a mixer for mixing the DC current signal and the periodic AC current signal to obtain a mixed current signal and for applying the mixed current signal to the capacitor microphone so that the capacitor microphone exhibits a limit cycle behavior.

[0007] In some implementations, the voltage amplitude of the DC current signal is less than 95% of a pull-in voltage amplitude of the capacitor microphone.

[0008] In some implementations, the voltage amplitude of the periodic AC current signal is less than 2% of a pull-in voltage amplitude of the capacitor microphone.

[0009] In some implementations, the voltage amplitude of the periodic AC current signal is greater than 0.5% of the pull-in voltage amplitude of the capacitor microphone.

[0010] In some implementations, a frequency of the periodic AC current signal is comprised between about 10% of a natural frequency of the membrane and the natural frequency of the capacitor microphone.

[0011] In some implementations, the AC signal generator comprises one of a Wien bridge circuit, a phase shift oscillator and a digital processor configured for direct digital synthesis.

[0012] In some implementations, the periodic AC current signal comprises a square wave signal.

[0013] In some implementations, the AC signal generator comprises a timer integrated circuit.

[0014] In some implementations, the periodic AC current signal comprises a triangular wave signal.

[0015] In some implementations, the AC signal generator comprises a Schmitt trigger and an Op-amp integrator circuit.

[0016] In accordance with another broad aspect, there is provided a method for extracting features from a sound time-series signal, the method comprising: generating a DC current signal, a voltage amplitude of the DC current signal being greater than an amplitude of a minimal operational bias voltage of the capacitor microphone, the capacitor microphone comprising a membrane and a backplate; generating a periodic AC current signal, a voltage amplitude of the periodic AC current signal being less than the voltage amplitude of the DC current signal; mixing together the DC current signal and the periodic AC current signal to obtain a mixed current signal; applying the mixed current signal to the capacitor microphone so that the capacitor microphone exhibits a limit cycle behavior; and detecting, by the capacitor microphone, the sound time-series signal and outputting, by the capacitor microphone, a feature signal indicative of features of data contained in the sound time-series signal.

[0017] In some implementations, the voltage of the DC current signal is less than 95% of the pull-in voltage of the capacitor microphone.

[0018] In some implementations, the voltage of the periodic AC current signal is less than 2% of the pull-in voltage of the capacitor microphone.

[0019] In some implementations, the voltage of the periodic AC current signal is greater than 0.5% of the pull-in voltage of the capacitor microphone.

[0020] In some implementations, a frequency of the periodic AC current signal is comprised between 10% of a natural frequency of the membrane and the natural frequency of the capacitor microphone.

[0021] In some implementations, the step of generating the periodic AC current signal is performed using one of a Wien bridge circuit, a phase shift oscillator and a digital processor configured for direct digital synthesis.

[0022] In some implementations, the periodic AC current signal comprises a square wave signal.

[0023] In some implementations, the step of generating the periodic AC current signal is performed using a timer integrated circuit.

[0024] In some implementations, the periodic AC current signal comprises a triangular wave signal.

[0025] In some implementations, the step of generating the periodic AC current signal is performed using a Schmitt trigger and an Op-amp integrator circuit.

[0026] In the following, the term “microphone” should be understood as referring to a condenser or capacitor microphone which is a type of microphones that relies on changes in capacitance for its operation. A condenser or capacitor microphone consists, inter alia, of a thin membrane in close proximity to a solid metal plate, as known in the art. The membrane acts as a diaphragm and is electrically conductive.

[0027] Implementations of the present technology each have at least one of the above-mentioned objects and / or aspects, but do not necessarily have all of them. It should be understood that some aspects of the present technology that have resulted from attempting to attain the above-mentioned object may not satisfy this object and / or may satisfy other objects not specifically recited herein.

[0028] Additional and / or alternative features, aspects and advantages of implementations of the present technology will become apparent from the following description, the accompanying drawings and the appended claims.BRIEF DESCRIPTION OF THE DRAWINGS(0029] For a beter understanding of the present technology, as well as other aspects and further features thereof, reference is made to the following description which is to be used in conjunction with the accompanying drawings, where:

[0001] Fig. 1 is a block diagram illustrating a system for analysing an audio signal, the system comprising a reservoir microphone provided with a backplate and a membrane, in accordance with an embodiment;

[0002] Fig. 2 schematically illustrates the backplate and the membrane of the microphone of Fig. 1, in accordance with an embodiment;

[0003] Fig. 3 illustrates an exemplary limit cycle behavior produced by an active condenser microphone reservoir, jiters in the curve being caused by sampling uncertainty;

[0004] Fig. 4 illustrates exemplary Fast Fourier transformations (FFT) of a signal output from a reservoir microphone with input tone at (a) 1 kHz, (b) 5 kHz (coincident with the activation frequency) and (c) 10 kHz, sample non-linear harmonics following Farey series being circled;

[0005] Fig. 5 illustrates an exemplary process for signal preprocessing to generate feature maps for audio recognition;

[0006] Figs. 6a, 6b and 6c illustrate exemplary feature maps generated from the algorithm of Fig. 5 when the detected audio signal corresponds to a dog barking, a siren and a jackhammer, respectively;

[0007] Figs. 7a - 7e illustrate exemplary feature maps of a same audio signal under different types of noise processed by reservoir microphone and generating the algorithm of Fig. 5 for the following cases: coqui sound with additional white noise in low signal to noise ratio (SNR), medium SNR and high SNR case, clean coqui sound and a rainy background mixed with the coqui foreground audio, respectively;

[0008] Fig. 8 illustrates exemplary audio recognition results using different machine learning models for a 4-class audio recognition task, in which the total length of the audio for training the models is 20 minutes (-100 pieces for each class);

[0009] Fig. 9 illustrates a setup using two moving components of a microphone to achieve hardware recovery of an original audio signal, in accordance with an embodiment;

[0010] Figs. 10a- lOd illustrate an exemplary waveform comparison to test the audio recovery methods in which Fig. 10a illustrates an original audio signal, Fig. 10b illustrates a direct output from the MEMS reservoir, Fig. 10c illustrates the output obtained using a software recovery method and Fig. lOd illustrates the output obtained using a hardware recovery method; and

[0011] Fig. 11 illustrates a flow chart of a method for analysing an audio signal, in accordance with one embodiment.DETAILED DESCRIPTION

[0030] The examples and conditional language recited herein are principally intended to aid the reader in understanding the principles of the present technology and not to limit its scope to such specifically recited examples and conditions. It will be appreciated that those skilled in the art may devise various arrangements which, although not explicitly described or shown herein, nonetheless embody the principles of the present technology and are included within its spirit and scope.

[0031] Furthermore, as an aid to understanding, the following description may describe relatively simplified implementations of the present technology. As persons skilled in the art would understand, various implementations of the present technology may be of a greater complexity.

[0032] In some cases, what are believed to be helpful examples of modifications to the present technology may also be set forth. This is done merely as an aid to understanding, and, again, not to define the scope or set forth the bounds of the present technology. These modifications are not an exhaustive list, and a person skilled in the art may make other modifications while nonetheless remaining within the scope of thepresent technology. Further, where no examples of modifications have been set forth, it should not be interpreted that no modifications are possible and / or that what is described is the sole manner of implementing that element of the present technology.

[0033] In the following, there is described a device and method for extracting features from au audio signal using the same microphone that is used for detecting the audio signal. The microphone is a capacitor or condenser microphone that includes a membrane in closed proximity to a backplate, as known in the art. A combination of AC and DC currents are applied to the backplate of the microphone so that the membrane oscillates and exhibits a limit cycle behavior, and the microphone acts an oscillation-based physical reservoir capable of extracting features from the detected audio signal.

[0034] Fig. 1 illustrates one embodiment of a system 10 for analysing an audio or sound signal such as identifying audio events within an audio time-series signal. In one embodiment, an audio event corresponds to an audio signature of an event. For example, an audio event may correspond to the audio signature of an action such as a vehicle passing -by and wake words.

[0035] The system 10 comprises a reservoir 12, an analog-to-digital (AD) converter 14 and a processor or processing unit 16. It should be understood that the signal outputted by the microphone is an analog signal which is converted into a digital signal by the AD converter 14.

[0036] The reservoir 12 is configured for detecting an audio time-series signal, extracting features from the data contained in the detected audio signal and outputting an analog feature signal indicative of the extracted features. The AD converter 14 is configured for receiving the analog feature signal and converting it into a digital feature signal which is also indicative of the extracted features. The processor 16 is configured for identifying at least one audio event from the digital feature signal and outputting a signal indicative of the identification of the audio event.

[0037] In one embodiment, the processor 18 is configured for executing a machine learning model for analysing the audio signal. Examples of adequate machine learning methods includes linear regression, decision tree, k-nearest neighboring, recurrent neural network, etc. In one embodiment, the machine learning model is trained prior tobe deployed to the processor 16. In this case, training audio segments or clips are fed into the feature extractor and the output signals of the feature extractor are used to train the machine learning model on a computer other than the processor 16, such as a server located in the cloud.

[0038] In one embodiment, the processor 16 is a lightweight processor such as a microcontroller.

[0039] Referring back to Fig. 1, the reservoir 12 comprises a capacitor or condenser microphone 20, a direct current (DC) generator source 22, an alternating current (AC) generator source 24 and a signal mixer 26. As illustrated in Fig. 2 and as known in the art, the capacitor or condenser microphone 20 comprises, inter alia, a backplate 30 and a membrane 32 positioned in closed proximity to the backplate 30. The DC current generator 22 is configured for generating a DC current signal and the AC current generator 24 is configured for generating an AC current signal. The signal mixer 26 is configured for receiving the DC current signal from the DC current generator 22 and the AC current signal from the AC current generator 24 and mixing together the DC and AC current signals to obtain a mixed current signal which corresponds to an addition of the DC current signal and the AC current signal.

[0040] In some implementations, the signal mixer 26 is configured for multiplying together the AC current signal and the DC current signal to obtain the mixed current signal. In other implementations, the signal mixer 26 is configured for adding together the AC current signal and the DC current signal to obtain the mixed current signal. In further implementations, the signal mixer 26 is configured for modulating the AC current signal and the DC current signal to obtain the mixed current signal.

[0041] The capacitor microphone 20 is configured for receiving the mixed current signal, which is applied to the backplate thereof, detecting the audio signal to be processed and extracting features from the detected audio signal when it is operated as described below.

[0042] It should be understood that the voltage amplitude of the DC current signal VDC is constant in time and the voltage amplitude of the AC current signal VAC is periodic.

[0043] The amplitude of the voltage VDC of the DC current signal is chosen so that the membrane 32 of the microphone 20 possesses high non-linearity, i.e., a small mechanical perturbation will generate unproportional oscillation of the membrane 32. In at least some embodiments, such high non-linearity is achieved by generating a DC current signal having a voltage amplitude VDC that is being greater than an amplitude of a minimal operational bias voltage of the microphone 20. In this case, the voltage amplitude VDC allows for generating an electrostatic force that is greater than the damping force in the microphone transducer.

[0044] In some embodiments, the voltage amplitude VDC of the DC current signal is less than about 95% of the pull-in or critical voltage amplitude of the microphone 20 to prevent the membrane 32 of the microphone 20 from being damaged. In one embodiment, the voltage amplitude VDC of the DC current signal is comprised between about 70% and about 95% of the pull-in voltage amplitude of the microphone 20.

[0045] The amplitude of the voltage VAC of the AC current signal is chosen to be less than that of the voltage VDC of the DC current signal. In some embodiments, the voltage amplitude VAC of the AC current signal is less than about 1% of the voltage amplitude VDC of the DC current signal. In the same or other embodiments, the voltage amplitude VAC of the AC current signal is less than about 2% of the pull-in voltage amplitude of the microphone 20. In some embodiments, the voltage amplitude VAC of the AC current signal is comprised between about 0.5% and about 2% of the pull-in voltage amplitude of the microphone 20. Such a voltage amplitude VAC applied on the backplate 30 of the microphone 20 allows for generating a mechanical vibration on the membrane 32 serving as the oscillatory activation for the audio signal.

[0046] In some embodiments, the frequency of the AC current signal is chosen so as to generate sufficient linear and non-linear harmonics covering from about 100 Hz to over 20 kHz required for audio processing while using small voltage amplitude for the AC current signal. In one embodiment, the frequency of the AC current signal is comprised between about 10% of the natural frequency of the membrane 32 and the natural frequency of the microphone 20, i.e., 100% of the natural frequency of the microphone 20. In some embodiments, the frequency of the AC current signal is equal to about one quarter of the natural frequency of the membrane 32.

[0047] The AC current signal generated by the AC current generator 24 may be any adequate periodic AC current signal having any adequate periodic shape. For example, the periodic signal generated by the AC current generator 24 may be a sinusoidal signal, a square wave signal, a triangular wave signal, or the like.

[0048] In one embodiment, the AC current generator 24 comprises a Wien bridge circuit configured for generating a periodic signal (such as a sinusoidal signal) having the above-described characteristics.

[0049] In another embodiment, the AC current generator 24 comprises an oscillator such as a phase shift oscillator configured for generating a periodic signal (such as a sinusoidal signal) having the above-described characteristics.

[0050] In a further embodiment, the AC current generator 24 comprises a digital processor configured for generating a digital periodic signal having the above-described characteristics by direct synthesis, and a digital-to-analog (DA) converter configured for converting the digital periodic signal into an analog periodic signal.

[0001] In an embodiment in which the periodic signal is a square wave signal, the AC current generator 24 may comprise a timer integrated signal such as a CMOS 555 integrated circuit or an equivalent and adequate integrated circuit.

[0052] In an embodiment in which the periodic signal comprises a triangular wave, the AC current generator 24 may comprise a Schmitt trigger and an Op-amp integrator circuit, or any adequate and equivalent integrated circuits.

[0053] It should be understood that any adequate DC current generator 22 may be used. For example, the DC current generator 22 may comprise a boost converter, a charge pump, an operational amplifier, or the like.

[0054] While the system 10 is described herein in relation with the identification of audio events, the person skilled in the art will understand that the system 10 may be configured for performing another task depending on how the processor 18 is configured. For example, the processor 18 may be configured for performing a task other than the identification of audio events from an audio time series signal, such as identifying voice biometrics, identifying audio signatures to predict anomalies ofindustrial equipment, etc. while using the same feature extractor as the one contained in the system 10.

[0055] It should be understood that the combination of the above-described voltages VDC and VAC to apply a combined voltage VDC + VAC to the backplate 32 of the microphone 20 allows for converting the assembly of the backplate 30 and the membrane 32 of the microphone 20 into a physical reservoir. The physical reservoir modulates the audio signal on the top of the oscillatory activation in a non-linear fashion to change instantaneously the limit cycle created by the oscillatory activation. The limit cycle occurring in the physical reservoir, is defined as that the two parameters in the non-linear system (i.e., the readout signal and its first derivative in this case) create a trajectory that will substantially never diverge as the time approaches infinity as illustrated in Fig. 3. As such, the limit cycle generates an attractor allowing a detected audio signal to be encoded and regularized on the attractor for machine perception substantially without any chaotic behavior. As a result, the above-described capacitor or condenser microphone may correspond to a reservoir microphone.

[0056] In the following, there is provided an experiment illustrating the high nonlinearity of the reservoir microphone (or capacitor microphone). Single tones with below, same, and higher than the frequency of AC current signal are played to the reservoir microphone and the outputs of these experiments illustrated show both linear and non-linear, as illustrated in Fig. 4, thereby demonstrating a broadband non-linear oscillation.

[0057] For an audio recognition task, the output signal from a reservoir microphone is directly fed to a low-level processor. Subsequently, the signal is processed by an algorithm such as the algorithm illustrated in Fig. 5 to generate a feature map for machine perception. The illustrated algorithm, by referencing the activation signal, rearranges the ID readout from the reservoir microphone to a 2D feature map for comparison of the radius change of the limit cycle. Followed by aZ-score normalization given by Eq. 1, the signal is processed to show the variations of signals over different cycles of the oscillation over time. In this equation, x corresponding to the original value of the signal, subtracted by the average of one oscillation cycle, this value is then divided by the standard deviation. The inverse tanh activation given by Eq. 2 is operated afterwards using the normalized values of signal such that the granularities of the audiosignal (e.g., the abrupt change of the volume and pitch in the audio) are further enhanced and random noise (usually low values around zero) is decreased or supressed. After inverse tanh activation, a delay is set to sample the data. After a sufficiently long period of time, the signals are ensemble averaged based on the cycle reading to further reduce the amount of data for machine learning readout. Last, the ensemble average reading of the oscillator signal is shifted to have universal zero-crossings (i.e., where the value of the reading changes from positive to negative) over different ensemble average readings of the signal to reduce the randomness in the processing.

[0058] As shown in Fig. 7, the rendering of the feature map shows the variations over different classes of the audios, perceivable by common audio recognition readouts. In addition, the features of the audio from reservoir are robust over a wide range of the additional white noise (Figs. 7a to 7c) and the mixer of the foreground audio and background audios (Figs. 7d and 7e). As such, the system built with the active condenser reservoir microphone could operate audio recognition substantially normally in noisy environments.

[0059] A number of machine learning algorithms (i.e., readouts) are constructed and tested for analyzing reservoir data. Before sending data to the readouts, an ensembled average of the 2D feature map over temporal direction is calculated to further reduce the amount of data to a 1 D vector for running recognition algorithm . Fig . 8 demonstrates that, for a 4-class urban audio recognition task, a machine learning model with less than -100 KB (i.e., operable on a low-level edge processor such as a microcontroller unit) could reach a high accuracy to around 96%.

[0060] Meanwhile, the original signal from the microphone is still recoverable with the modification of the power circuit. Fig. 9 shows a conceptual design of using two moving components 50 and 52 (such as two cantilevers) oscillating out-of-phase to cancel out the reservoir effects as the hardware method to recover the original audio. This recovery is also achievable numerically (i.e., using software) by detecting datapoints sample out-of-phase from the MEMS microphone with just one moving component.

[0061] As depicted in Fig. 10, the effects ofthe reservoir microphone (shown in Fig. 10b) on the detected or original audio signal (illustrated in Fig. 10a) could be reversed by both software (Fig. 10c) and hardware (Fig. lOd) approaches and the recovered signals largely resemble the original audio signal, thereby demonstrating the recovery of the original audio from the reservoir microphone which is more similar to a gapclosing actuator instead of a microphone operating normally.

[0062] Fig. 11 illustrates one embodiment of a method 100 for extracting features from a sound or audio time-series signal. The method 100 is executed using a capacitor microphone comprising at least a membrane and a backplate, such as capacitor microphone 20.

[0063] At step 102, a DC current signal having a voltage amplitude being greater than the amplitude of the minimal operational bias voltage of the capacitor microphone is generated. As described above, any adequate DC current generator may be used at step 102.

[0064] At step 104, a periodic AC current signal having a voltage amplitude that is less than the voltage amplitude of the DC current signal is generated. As described above, any adequate AC current generator configured for generating a periodic AC current signal may be used at step 104.

[0065] At step 106, the DC current signal and the periodic AC current signal are mixed together, thereby obtaining a mixed current signal. As described above, any adequate mixing method may be used at step 106. For example, the AC current signal and the DC current signal may be multiplied together at step 106 to obtain the mixed current signal. In another example, the AC current signal and the DC current signal may be added together at step 106 to obtain the mixed current signal. In a further example, the AC current signal and the DC current signal may be modulated at step 106 to obtain the mixed current signal.

[0066] At step 108, the mixed current signal obtained at step 106 is applied to the capacitor microphone so that the capacitor microphone exhibits a limit cycle behavior.

[0067] At step 110, a sound or audio time-series signal is detected by the capacitor microphone (while exhibiting the limit cycle behavior). As a result of the detection of the sound time-series signal, the capacitor microphone outputs a feature signal indicative of features of data contained in the sound time-series signal.

[0068] In some implementations, the voltage of the DC current signal is chosen to be less than 95% of the pull-in voltage of the capacitor microphone.

[0069] In some implementations, the voltage of the periodic AC current signal is less than 2% of the pull-in voltage of the capacitor microphone.

[0070] In some implementations, the voltage of the periodic AC current signal is greater than 0.5% of the pull-in voltage of the capacitor microphone.

[0071] In some implementations, the frequency of the periodic AC current signal is comprised between about 10% of the natural frequency of the membrane and the natural frequency of the capacitor microphone.

[0072] In some implementations, the generation of the periodic AC current signal is performed using a Wien bridge circuit, a phase shift oscillator or a digital processor configured for direct digital synthesis.

[0073] In some implementations, the periodic AC current signal comprises a square wave signal.

[0074] In some implementations, the generation of the periodic AC current signal is performed using a timer integrated circuit.

[0075] In some implementations, the periodic AC current signal comprises a triangular wave signal.

[0076] In some implementations, the generation of the periodic AC current signal is performed using a Schmitt trigger and an Op-amp integrator circuit.

[0077] In some embodiments, the present method and system allow for carrying out on-the-edge audio recognition using machine learning. The audio recognition system built on the present technology is reconfigurable and eliminates the need for intermediate hardware and processing steps including digital signal processing chip,digital to analog conversion, dedicated Al chip, and / or data preprocessing such as Mel spectrogram. For end users, the construction of such system is easy with only a microphone and a low-level microprocessor, or the microphone can be seamlessly integrated to the existing audio processing pipeline by upgrading a few MEMS devices. For OEMs, the present system and method add new functionality (e.g., audio signal recognition) to a microphone without sacrificing audio quality with existing audio readout hardware such that an OEM may open new markets and paradigms by employing new hardware with limited redesign fabricated in existing facilities.

[0078] In some embodiment, the present technology may be paired with a low-level microprocessor designed for industrial and institutional use cases such as predictive maintenance and vehicle tracking using audios.

[0079] In some embodiments, the present technology may serve as a MEMS meaning sensor integrated to an existing consumer electronics for human related voice and speech recognition (e.g., wake words, soft biometric, and voice command and control).

[0080] Modifications and improvements to the above-described implementations of the present technology may become apparent to those skilled in the art. The foregoing description is intended to be exemplary rather than limiting.

Claims

CLAIMSWhat is claimed is:

1. A device for extracting features from a sound time-series signal, the device comprising: a capacitor microphone comprising a membrane and a backplate, the capacitor microphone for detecting the sound time-series signal and outputting a feature signal indicative of features of data contained in the sound time-series signal; a DC signal generator for generating a DC current signal, a voltage amplitude of the DC current signal being greater than an amplitude of a minimal operational bias voltage of the capacitor microphone; an AC signal generator for generating a periodic AC current signal, a voltage amplitude of the periodic AC current signal being less than the voltage amplitude of the DC current signal; and a mixer for mixing the DC current signal and the periodic AC current signal to obtain a mixed current signal and for applying the mixed current signal to the capacitor microphone so that the capacitor microphone exhibits a limit cycle behavior.

2. The device of claim 1, wherein the voltage amplitude of the DC current signal is less than 95% of a pull-in voltage amplitude of the capacitor microphone.

3. The device of claim 1 , wherein the voltage amplitude of the periodic AC current signal is less than 2% of a pull-in voltage amplitude of the capacitor microphone.

4. The device of claim 3, wherein the voltage amplitude of the periodic AC current signal is greater than 0.5% of the pull-in voltage amplitude of the capacitor microphone.

5. The device of any one of claims 1 to 4, wherein a frequency of the periodic AC current signal is comprised between about 10% of a natural frequency of the membrane and the natural frequency of the capacitor microphone.

6. The device of any one of claims 1 to 5, wherein the AC signal generator comprises one of a Wien bridge circuit, a phase shift oscillator and a digital processor configured for direct digital synthesis.

7. The device of any one of claims 1 to 6. wherein the periodic AC current signal comprises a square wave signal.

8. The device of claim 7, wherein the AC signal generator comprises a timer integrated circuit.

9. The device of any one of claims 1 to 6, wherein the periodic AC current signal comprises a triangular wave signal.

10. The device of claim 9, wherein the AC signal generator comprises a Schmitt trigger and an Op-amp integrator circuit.

11. A method for extracting features from a sound time-series signal, the method comprising: generating a DC current signal, a voltage amplitude of the DC current signal being greater than an amplitude of a minimal operational bias voltage of the capacitor microphone, the capacitor microphone comprising a membrane and a backplate; generating a periodic AC current signal, a voltage amplitude of the periodic AC current signal being less than the voltage amplitude of the DC current signal; mixing together the DC current signal and the periodic AC current signal to obtain a mixed current signal; applying the mixed current signal to the capacitor microphone so that the capacitor microphone exhibits a limit cycle behavior; and detecting, by the capacitor microphone, the sound time-series signal and outputting, by the capacitor microphone, a feature signal indicative of features of data contained in the sound time-series signal.

12. The method of claim 11, wherein the voltage of the DC current signal is less than 95% of the pull-in voltage of the capacitor microphone.

13. The method of claim 11 or 12, wherein the voltage of the periodic AC current signal is less than 2% of the pull-in voltage of the capacitor microphone.

14. The method of claim 13, wherein the voltage of the periodic AC current signal is greater than 0.5% of the pull-in voltage of the capacitor microphone.

15. The method of any one of claims 11 to 14, wherein a frequency of the periodic AC current signal is comprised between 10% of a natural frequency of the membrane and the natural frequency of the capacitor microphone.

16. The method of any one of claims 11 to 15, wherein said generating the periodic AC current signal is performed using one of a Wien bridge circuit, a phase shift oscillator and a digital processor configured for direct digital synthesis.

17. The method of any one of claims 11 to 16, wherein the periodic AC current signal comprises a square wave signal.

18. The method of claim 17, wherein said generating the periodic AC current signal is performed using a timer integrated circuit.

19. The method of any one of claims 11 to 16, wherein the periodic AC current signal comprises a triangular wave signal.

20. The method of claim 19, wherein said generating the periodic AC current signal is performed using a Schmitt trigger and an Op-amp integrator circuit.

Images