Abstract: Audio beamforming systems and methods that enable more precise control of lobes and nulls of an array microphone are provided. Optimized beamformer coefficients can be generated to result in beamformed signals associated with one or more lobes steered towards one or more desired sound locations and one or more nulls steered towards one or more undesired sound location. The performance of acoustic echo cancellation can be improved and enhanced.

Abstract: Techniques are disclosed herein for providing front-end audio processing for automatic speech recognition. Examples may include receiving audio data from one or more microphone array devices located within an audio environment, extracting an audio feature set from the audio data, inputting the audio feature set to an audio source separation model to generate an audio speech signal configured for automatic speech recognition (ASR), inputting the audio speech signal to an ASR model configured to generate textual data, and outputting the textual data to a post-processing system.

Abstract: Techniques are disclosed herein for providing beamforming for at least one microphone array based at least in part on a steered response power (SRP) transformation of audio data. Examples may include receiving audio data from multiple audio capture devices comprising at least one microphone array located within an audio environment. Examples may also include generating an SRP transformation of the audio data. The SRP transformation may comprise a set of SRP weights for a spatial coordinate grid representing the audio environment. Examples may also include performing, based at least in part on a signal-to-noise ratio (SNR) estimate associated with the SRP transformation, one or more of beamforming steering or beamforming selection with respect to the at least one linear array microphone.

Abstract: Techniques are disclosed herein for providing full-band audio signal reconstruction enabled by output from a machine learning model trained based on an audio feature set extracted from a portion of the audio signal. Examples may include generating a model input audio feature set for a first frequency portion of an audio signal defined based on a hybrid audio processing frequency threshold. Examples may also include inputting the model input audio feature set to a machine learning model configured to generate a frequency characteristics output related to the first frequency portion of the audio signal. Examples may also include applying the frequency characteristics output to at least a second frequency portion of the audio signal to generate a reconstructed full-band audio signal.

Abstract: Hybrid audio beamforming systems and methods with narrower beams and improved directivity are provided. The hybrid audio beamforming system includes a time domain beamformer for processing upper frequency band signals of an audio signal using a time domain beamforming technique, and a frequency domain beamformer for processing groups of lower frequency band signals of the audio signal using frequency domain beamforming techniques.

Abstract: Audio beamforming systems and methods that enable more precise control of lobes and nulls of an array microphone are provided. Optimized beamformer coefficients can be generated to result in beamformed signals associated with one or more lobes steered towards one or more desired sound locations and one or more nulls steered towards one or more undesired sound location. The performance of acoustic echo cancellation can be improved and enhanced.

Abstract: Systems and methods are disclosed for providing voice and noise activity detection with audio automixers that can reject errant non-voice or non-human noises while maximizing signal-to-noise ratio and minimizing audio latency.

Abstract: A wireless audio system for encoding and decoding an audio signal using spectral bandwidth replication is provided. Bandwidth extension is performed in the time-domain, enabling low-latency audio coding.

Abstract: An audio signal processor includes a digital filter block configured to receive an audio signal and output a first filtered audio signal, and a phase linearization block configured to receive the first filtered audio signal and output a second filtered audio signal with a more linear phase.

Abstract: Systems and methods are disclosed for providing voice and noise activity detection with audio automixers that can reject errant non-voice or non-human noises while maximizing signal-to-noise ratio and minimizing audio latency.

Abstract: Hybrid audio beamforming systems and methods with narrower beams and improved directivity are provided. The hybrid audio beamforming system includes a time domain beamformer for processing upper frequency band signals of an audio signal using a time domain beamforming technique, and a frequency domain beamformer for processing groups of lower frequency band signals of the audio signal using frequency domain beamforming techniques.

Abstract: Systems and methods are disclosed for providing voice and noise activity detection with audio automixers that can reject errant non-voice or non-human noises while maximizing signal-to-noise ratio and minimizing audio latency.

Abstract: A wireless audio system for encoding and decoding an audio signal using spectral bandwidth replication is provided. Bandwidth extension is performed in the time-domain, enabling low-latency audio coding.

Abstract: An audio signal processor includes a digital filter block configured to receive an audio signal and output a first filtered audio signal, and a phase linearization block configured to receive the first filtered audio signal and output a second filtered audio signal with a more linear phase.

Abstract: A wireless audio system for encoding and decoding an audio signal using spectral bandwidth replication is provided. Bandwidth extension is performed in the time-domain, enabling low-latency audio coding.

Abstract: Systems and methods are disclosed for providing voice and noise activity detection with audio automixers that can reject errant non-voice or non-human noises while maximizing signal-to-noise ratio and minimizing audio latency.

Abstract: A method for long impulse response digital filtering of an input data stream, by use of a digital filtering system. Where the input data stream is divided into zero-input signals and zero-state signals. One of the zero-input signals and a corresponding impulse response of the digital filtering system is converted to the frequency domain to determine a respective zero-input response of the digital filtering system. One of the zero-state signals is convolved with a corresponding impulse response of the digital filtering system to determine a respective zero-state response of the digital filtering system, wherein at least part of the zero-input signal precedes the zero-state signal. The zero-state response of the digital filtering system is added to the zero-input response of the digital filtering system to determine the response of the digital filtering system. Apparatus for effecting this method is also disclosed.

Abstract: A method for long impulse response digital filtering of an input data stream, by use of a digital filtering system. Where the input data stream is divided into zero-input signals and zero-state signals. One of the zero-input signals and a corresponding impulse response of the digital filtering system is converted lo the frequency domain to determine a respective zero-input response of the digital filtering system. One of the zero-state signals is convolved with a corresponding impulse response of the digital filtering system to determine a respective zero-state response of the digital filtering system, wherein at least part of the zero-input signal precedes the zero-stale signal. The zero-state response of the digital filtering system is added to the zero-input response or the digital filtering system to determine the response of the digital filtering system. Apparatus for effecting this method is also disclosed.

Abstract: A system and method for determining DTMF coefficients stores a plurality of reference templates, at least some of said reference templates representative of likelihood ratios. A plurality of LPC coefficients is received. The LPC coefficients are representative of an input signal. The system and method determines a plurality of current likelihood ratios based, at least in part, upon the LPC coefficients. The system and method also determines an initial tone based, at least in part, on the minimum of the current likelihood ratios and the minimum of the reference likelihood ratios. A plurality of LSF coefficients is received. The LSF coefficients are determined, at least in part, upon the input signal. The system and method verifies the validity of the initial tone based, at least in part, upon the LSF coefficients.

Abstract: A system for creating a comfort noise signal, said system includes a clipping circuit, a comfort noise generator, and a switch. The clipping circuit receives a near-end signal, a far-end signal and a residual signal. The clipping circuit determines the power of the far-end signal, the power of the near-end signal and the power of the residual signal. The clipping circuit determines the ratio of the power of the near-end signal to the residual signal and whether the ratio is less than a predetermined threshold. A comfort noise determination block is coupled to the residual signal and selectively forms a comfort noise signal. A switch is coupled to the comfort noise determination block and the clipping circuit. The switch is activated to transmit the comfort noise signal whenever the ratio of the power of the near-end signal to the residual signal is less than a predetermined threshold.