Abstract: A method of operation of a device includes receiving an input signal at the device. The input signal is generated using at least one microphone. The input signal includes a first signal component having a first amount of wind turbulence noise and a second signal component having a second amount of wind turbulence noise that is greater than the first amount of wind turbulence noise. The method further includes generating, based on the input signal, an output signal at the device. The output signal includes the first signal component and a third signal component that replaces the second signal component. A first frequency response of the input signal corresponds to a second frequency response of the output signal.

Abstract: An apparatus for detecting a sound in an acoustical environment includes a microphone array configured to detect an audio signal in the acoustical environment. The apparatus also includes a processor configured to determine an angular location of a sound source of the audio signal. The angular location is relative to the microphone array. The processor is also configured to determine at least one reverberation characteristic of the audio signal. The processor is further configured to determine a distance, relative to the microphone array, of the sound source along an axis associated with the angular location based on the at least one reverberation characteristic.

Abstract: A device includes a memory and a processor. The memory is configured to store a threshold. The processor is configured to authenticate a user based on authentication data. The processor is also configured to, in response to determining that the user is authenticated, generate a correlation score indicating a correlation between a first signal received from a first sensor and a second signal received from a second sensor. The processor is also configured to determine liveness of the user based on a comparison of the correlation score and the threshold.

Abstract: Disclosed is a feature extraction and classification methodology wherein audio data is gathered in a target environment under varying conditions. From this collected data, corresponding features are extracted, labeled with appropriate filters (e.g., audio event descriptions), and used for training deep neural networks (DNNs) to extract underlying target audio events from unlabeled training data. Once trained, these DNNs are used to predict underlying events in noisy audio to extract therefrom features that enable the separation of the underlying audio events from the noisy components thereof.

Abstract: A method for speech restoration by an electronic device is described. The method includes obtaining a noisy speech signal. The method also includes suppressing noise in the noisy speech signal to produce a noise-suppressed speech signal. The noise-suppressed speech signal has a bandwidth that includes at least three subbands. The method further includes iteratively restoring each of the at least three subbands. Each of the at least three subbands is restored based on all previously restored subbands of the at least three subbands.

Abstract: Disclosed is a feature extraction and classification methodology wherein audio data is gathered in a target environment under varying conditions. From this collected data, corresponding features are extracted, labeled with appropriate filters (e.g., audio event descriptions), and used for training deep neural networks (DNNs) to extract underlying target audio events from unlabeled training data. Once trained, these DNNs are used to predict underlying events in noisy audio to extract therefrom features that enable the separation of the underlying audio events from the noisy components thereof.

Abstract: A method for speech restoration by an electronic device is described. The method includes obtaining a noisy speech signal. The method also includes suppressing noise in the noisy speech signal to produce a noise-suppressed speech signal. The noise-suppressed speech signal has a bandwidth that includes at least three subbands. The method further includes iteratively restoring each of the at least three subbands. Each of the at least three subbands is restored based on all previously restored subbands of the at least three subbands.

Abstract: A method for signal level matching by an electronic device is described. The method includes capturing a plurality of audio signals from a plurality of microphones. The method also includes determining a difference signal based on an inter-microphone subtraction. The difference signal includes multiple harmonics. The method also includes determining whether a harmonicity of the difference signal exceeds a harmonicity threshold. The method also includes preserving the harmonics to determine an envelope. The method further applies the envelope to a noise-suppressed signal.

Abstract: Systems, methods, and apparatus for pitch trajectory analysis are described. Such techniques may be used to remove vocals and/or vibrato from an audio mixture signal. For example, such a technique may be used to pre-process the signal before an operation to decompose the mixture signal into individual instrument components.

Abstract: A device includes a memory, a receiver, a processor, and a display. The memory is configured to store a speaker model. The receiver is configured to receive an input audio signal. The processor is configured to determine a first confidence level associated with a first portion of the input audio signal based on the speaker model. The processor is also configured to determine a second confidence level associated with a second portion of the input audio signal based on the speaker model. The display is configured to present a graphical user interface associated with the first confidence level or associated with the second confidence level.

Abstract: A method for speech modeling by an electronic device is described. The method includes obtaining a real-time noise reference based on a noisy speech signal. The method also includes obtaining a real-time noise dictionary based on the real-time noise reference. The method further includes obtaining a first speech dictionary and a second speech dictionary. The method additionally includes reducing residual noise based on the real-time noise dictionary and the first speech dictionary to produce a residual noise-suppressed speech signal at a first modeling stage. The method also includes generating a reconstructed speech signal based on the residual noise-suppressed speech signal and the second speech dictionary at a second modeling stage.

Abstract: A method for enhancing an audio signal by an electronic device is described. The method includes determining formant peaks based on an audio signal. The method also includes generating formant peak models. Generating formant peak models includes individually modeling each formant peak. The method further includes generating a global envelope based on the formant peak models.

Abstract: Decomposition of a multi-source signal using a basis function inventory and a sparse recovery technique is disclosed.

Abstract: Systems, methods, and apparatus for pitch trajectory analysis are described. Such techniques may be used to remove vocals and/or vibrato from an audio mixture signal. For example, such a technique may be used to pre-process the signal before an operation to decompose the mixture signal into individual instrument components.

Abstract: A method for detecting voice activity by an electronic device is described. The method includes detecting near end speech based on a near end voiced speech detector and at least one single channel voice activity detector. The near end voiced speech detector is associated with a harmonic statistic based on a speech pitch histogram.

Abstract: A method for signal level matching by an electronic device is described. The method includes capturing a plurality of audio signals from a plurality of microphones. The method also includes determining a difference signal based on an inter-microphone subtraction. The difference signal includes multiple harmonics. The method also includes determining whether a harmonicity of the difference signal exceeds a harmonicity threshold. The method also includes preserving the harmonics to determine an envelope. The method further applies the envelope to a noise-suppressed signal.

Abstract: A method for restoring a processed speech signal by an electronic device is described. The method includes obtaining at least one audio signal. The method also includes performing bin-wise voice activity detection based on the at least one audio signal. The method further includes restoring the processed speech signal based on the bin-wise voice activity detection.

Abstract: Decomposition of a multi-source signal using a basis function inventory and a sparse recovery technique is disclosed.