Abstract: A method of detecting a suitability of a signal for live speech detection, the method comprising: receiving the signal containing speech from a transducer; measuring a signal characteristic of an audible component of the received signal; estimating an expected signal characteristic of an ultrasonic component of the received signal based on the measured signal characteristic of the audible component; determining, based on the estimated expected signal characteristic, whether the ultrasonic component is suitable for detecting whether the speech is live speech.

Abstract: An estimator of direction of arrival (DOA) of speech from a far-field talker to a device in the presence of room reverberation and directional noise includes audio inputs received from multiple microphones and one or more beamformer outputs generated by processing the microphone inputs. A first DOA estimate is obtained by performing generalized cross-correlation between two or more of the microphone inputs. A second DOA estimate is obtained by performing generalized cross-correlation between one of the one or more beamformer outputs and one or more of: the microphone inputs and other of the one or more beamformer outputs. A selector selects the first or second DOA estimate based on an SNR estimate at the microphone inputs and a noise reduction amount estimate at the beamformer outputs. The SNR and noise reduction estimates may be obtained based on the detection of a keyword spoken by a desired talker.

Abstract: A system performs pole-zero or IIR modeling and estimation of an inter-microphone transfer function between first and second microphones that output respective first and second microphone signals. The system includes a first adaptive FIR filter to which the first microphone signal is provided, a delay element that delays the second microphone signal by a predetermined delay amount, and a second adaptive FIR filter to which the delayed second microphone signal is provided. A first coefficient of the second adaptive FIR filter is constrained to a fixed non-zero value. The filters are jointly adapted to minimize an error signal that is a difference of the two filters outputs. The delay is small: approximately the acoustic propagation delay between the two microphones and is not determined by the environmental reverberation characteristics. The error signal may serve as a noise reference in a noise canceller, for implementing far-field beamforming with low delay.

Abstract: A multi-microphone algorithm for detecting and differentiating interference sources from desired talker speech in advanced audio processing for smart home applications is described. The approach is based on characterizing a persistent interference source when sounds repeated occur from a fixed spatial location relative to the device, which is also fixed. Some examples of such interference sources include TV, music system, air-conditioner, washing machine, and dishwasher. Real human talkers, in contrast, are not expected to remain stationary and speak continuously from the same position for a long time. The persistency of an acoustic source is established based on identifying historically-recurring inter-microphone frequency-dependent phase profiles in multiple time periods of the audio data. The detection algorithm can be used with a beamforming processor to suppress the interference and for achieving voice quality and automatic speech recognition rate improvements in smart home applications.

Abstract: A system performs pole-zero or IIR modeling and estimation of an inter-microphone transfer function between first and second microphones that output respective first and second microphone signals. The system includes a first adaptive FIR filter to which the first microphone signal is provided, a delay element that delays the second microphone signal by a predetermined delay amount, and a second adaptive FIR filter to which the delayed second microphone signal is provided. A first coefficient of the second adaptive FIR filter is constrained to a fixed non-zero value. The filters are jointly adapted to minimize an error signal that is a difference of the two filters outputs. The delay is small: approximately the acoustic propagation delay between the two microphones and is not determined by the environmental reverberation characteristics. The error signal may serve as a noise reference in a noise canceller, for implementing far-field beamforming with low delay.

Abstract: An estimator of direction of arrival (DOA) of speech from a far-field talker to a device in the presence of room reverberation and directional noise includes audio inputs received from multiple microphones and one or more beamformer outputs generated by processing the microphone inputs. A first DOA estimate is obtained by performing generalized cross-correlation between two or more of the microphone inputs. A second DOA estimate is obtained by performing generalized cross-correlation between one of the one or more beamformer outputs and one or more of: the microphone inputs and other of the one or more beamformer outputs. A selector selects the first or second DOA estimate based on an SNR estimate at the microphone inputs and a noise reduction amount estimate at the beamformer outputs. The SNR and noise reduction estimates may be obtained based on the detection of a keyword spoken by a desired talker.

Abstract: An acoustic echo path change detector provides a monitoring process in an acoustic echo canceler that removes echo from a microphone signal using an adaptive echo path model that generates an echo estimate from a playback signal. The acoustic echo canceler removes the echo estimate from the microphone signal to provide an echo-canceled output signal. The path change detector receives the microphone signal, the echo estimate and the output signal and determines a rate of change of one or more statistical values dependent on the microphone signal, the echo estimate and the output signal. If the rate of change exceeds a threshold value, the echo path change detector generates an indication that causes a supervisory process to change adaptation of the adaptive echo path model to increase responsiveness to the change in the acoustic echo path, e.g., by increasing the step size.

Abstract: A method and apparatus to determine a direction of arrival (DOA) of a talker in the presence of a source of spatially-coherent noise. A time sequence of audio samples that include the spatially-coherent noise is received and buffered. Aided by previously known data, a trigger point is detected in the time sequence of audio samples when the talker begins to talk. The buffered time sequence of audio samples is separated into a noise segment and a signal-plus-noise segment based on the trigger point. For each direction of a plurality of distinct directions: an energy difference is computed for the direction between the noise segment and the signal-plus-noise segment, and the DOA of the talker is selected as the direction of the plurality of distinct directions having a largest of the computed energy differences.

Abstract: The reliable differentiation of human and artificial talkers is important for many automatic speaker verification applications, such as in developing anti-spoofing countermeasures against replay attacks for voice biometric authentication. A multi-microphone approach may exploit small movements of human talkers to differentiate between a human talker and an artificial talker. One method of determining the presence or absence of talker movement includes monitoring the variation of the inter-mic frequency-dependent phase profile of the received microphone array data over a period of time. Using spatial information with spectral-based techniques for determining whether an audio source is a human or artificial talker may reduce the likelihood of success of spoofing attacks against a voice biometric authentication system. The anti-spoofing countermeasure may be used in electronic devices including smart home devices, cellular phones, tablets, and personal computers.

Abstract: A method and apparatus to determine a direction of arrival (DOA) of a talker in the presence of a source of spatially-coherent noise. A time sequence of audio samples that include the spatially-coherent noise is received and buffered. Aided by previously known data, a trigger point is detected in the time sequence of audio samples when the talker begins to talk. The buffered time sequence of audio samples is separated into a noise segment and a signal-plus-noise segment based on the trigger point. For each direction of a plurality of distinct directions: an energy difference is computed for the direction between the noise segment and the signal-plus-noise segment, and the DOA of the talker is selected as the direction of the plurality of distinct directions having a largest of the computed energy differences.

Abstract: A change in the phase pattern of the inter-mic impulse response (IMIR), determined by a cross power spectral density, may be used to detect the appearance of a new talker or a dramatic movement of the current talker. For example, the phase of the IMIR is dependent on a location of the sound source relative to the microphone array. Any signal originating from a specific location has a specific phase pattern on the IMIR across the frequency domain. By comparing phase patterns of the current cross power spectral density with a recorded talker phase profile, a talker change can be detected. This detection can be used to control signal processing algorithms. For example, when talker change is detected, the step size of an adaptive filter can be increased to track the changes efficiently.

Abstract: The reliable differentiation of human and artificial talkers is important for many automatic speaker verification applications, such as in developing anti-spoofing countermeasures against replay attacks for voice biometric authentication. A multi-microphone approach may exploit small movements of human talkers to differentiate between a human talker and an artificial talker. One method of determining the presence or absence of talker movement includes monitoring the variation of the inter-mic frequency-dependent phase profile of the received microphone array data over a period of time. Using spatial information with spectral-based techniques for determining whether an audio source is a human or artificial talker may reduce the likelihood of success of spoofing attacks against a voice biometric authentication system. The anti-spoofing countermeasure may be used in electronic devices including smart home devices, cellular phones, tablets, and personal computers.

Abstract: A change in the phase pattern of the inter-mic impulse response (IMIR), determined by a cross power spectral density, may be used to detect the appearance of a new talker or a dramatic movement of the current talker. For example, the phase of the IMIR is dependent on a location of the sound source relative to the microphone array. Any signal originating from a specific location has a specific phase pattern on the IMIR across the frequency domain. By comparing phase patterns of the current cross power spectral density with a recorded talker phase profile, a talker change can be detected. This detection can be used to control signal processing algorithms. For example, when talker change is detected, the step size of an adaptive filter can be increased to track the changes efficiently.

Abstract: A multi-microphone algorithm for detecting and differentiating interference sources from desired talker speech in advanced audio processing for smart home applications is described. The approach is based on characterizing a persistent interference source when sounds repeated occur from a fixed spatial location relative to the device, which is also fixed. Some examples of such interference sources include TV, music system, air-conditioner, washing machine, and dishwasher. Real human talkers, in contrast, are not expected to remain stationary and speak continuously from the same position for a long time. The persistency of an acoustic source is established based on identifying historically-recurring inter-microphone frequency-dependent phase profiles in multiple time periods of the audio data. The detection algorithm can be used with a beamforming processor to suppress the interference and for achieving voice quality and automatic speech recognition rate improvements in smart home applications.

Abstract: A method and apparatus for audio privacy may be based on user identification. An audio signal containing speech may be analyzed, identifying a user to which the speech belongs and determining a user class for the user. The speech may be uploaded to a remote device based on whether the user class for the user is a public user class or a private user class. This allows certain users to opt-out of having their speech uploaded through public networks. The user identification may be based on voice biometrics.

Abstract: Noise sources may be identified as either an interference source, such as a television, or a talker source by analyzing phase information of the microphone signals. A phase delay variance may be computed from pairs of microphone signals. A profile of an interference source may be learned over time by updating a stored profile when the phase delay variance is below a threshold. The stored profile may be used to identify interference sources received by the microphones by determining a correlation between the microphone signals and the stored profile. When an interference source is detected, control parameters may be generated to control a beamformer to reduce contribution of the interference source to an output audio signal. The output audio signal may be used for speech processing, such as in a smart home device.

Abstract: Test apparatus measuring relative frequency response of first and second microphones includes a rotatable carrier. First and second microphones are sealingly clamped against a mounting surface of the carrier aligned with first and second apertures therein, such apertures lying equidistant from, and on opposite sides of, the carrier's axis of rotation. The carrier initially positions the first microphone closest to an audible signal source, and the responses of the microphones to an audible excitation signal are measured. The carrier is rotated 180 degrees, and the measurements are repeated. Elongated strips of gasket material are used to align the microphones and to form a seal with the carrier. When microphones are mounted deep within an audio device, the audio device is sealingly clamped against a mounting plate, sequentially aligning the mounting plate aperture with first and second apertures of the audio device housing corresponding to first and second microphones disposed therein.

Abstract: Test apparatus measuring relative frequency response of first and second microphones includes a rotatable carrier. First and second microphones are sealingly clamped against a mounting surface of the carrier aligned with first and second apertures therein, such apertures lying equidistant from, and on opposite sides of, the carrier's axis of rotation. The carrier initially positions the first microphone closest to an audible signal source, and the responses of the microphones to an audible excitation signal are measured. The carrier is rotated 180 degrees, and the measurements are repeated. Elongated strips of gasket material are used to align the microphones and to form a seal with the carrier. When microphones are mounted deep within an audio device, the audio device is sealingly clamped against a mounting plate, sequentially aligning the mounting plate aperture with first and second apertures of the audio device housing corresponding to first and second microphones disposed therein.

Abstract: Test apparatus measuring relative frequency response of first and second microphones includes a rotatable carrier. First and second microphones are sealingly clamped against a mounting surface of the carrier aligned with first and second apertures therein, such apertures lying equidistant from, and on opposite sides of, the carrier's axis of rotation. The carrier initially positions the first microphone closest to an audible signal source, and the responses of the microphones to an audible excitation signal are measured. The carrier is rotated 180 degrees, and the measurements are repeated. Elongated strips of gasket material are used to align the microphones and to form a seal with the carrier. When microphones are mounted deep within an audio device, the audio device is sealingly clamped against a mounting plate, sequentially aligning the mounting plate aperture with first and second apertures of the audio device housing corresponding to first and second microphones disposed therein.

Abstract: A telephone includes a transmit channel and a receive channel, each including a bank of sub-band filters having a VAD coupled one to each sub-band filter. Each VAD measures the spectral energy in a sub-band, compares the spectral energy to a first threshold, and produces an output signal representative of whether or not the first threshold is exceeded. The voice activity detector also includes a threshold circuit for calculating a dynamically adjustable noise threshold based upon averaged measured spectral energy. A wide band or system VAD monitors echo canceling circuitry to detect voice activity and double talk.