Mindspeed Technologies, Inc.

Conexant Systems, Inc.

Aondevices, Inc.

Rockwell International Corporation

Nytell Software LLC

Wiav Solutions LLC

Htc Corporation

Adil Benyassine

Eyal Shlomot

Huan-Yu Su

Yang Gao

Jes Thyssen

Mouna Elkhatib

Carlo Murgia

Jeffrey D Klein

Nick J Lavrov

Patrick D Ryan

Aruna Vittal

Eric Kwok Fung Yuen

Eric Kwok Yuen

Daniel Schoch

Eli Uc

Systems and methods are presented herein that include utilizing a specialized circuit architecture for the purpose of locating lost items through voice command recognition. The system can be customized with minimal, low power components to continuously monitor for a locating phrase and produce an audible alert when the locating phrase is detected. The system can be included in a stand-alone portable device that can be attached to electronic and non-electronic portable objects, or the system can be embedded with other circuitry into a portable electronic device. The system can be powered by battery such that it is practical for the device to function as a locating device, continuously monitoring for the locating phrase. Voice recognition can be performed completely within the system, providing the benefit of privacy.

System architecture and embedded circuit to locate a lost portable device using voice command

A multi-stage noise suppression system for reducing noise components in a noisy input signal has a first stage neural network that estimates a noise power spectrum for the noisy input signal. A first set of gain values corresponding to the noise power spectrum is generated by the first stage neural network. A second stage neural network estimates clean signal power spectrum values, which are derived from an application of a second set of gain values generated as a function of the clean signal power spectrum values and a first stage reduced noise signal power spectrum values.

Multistage low power, low latency, and real-time deep learning single microphone noise suppression

Systems and methods are presented for recognizing and responding to voice commands at a local system and selectively streaming audio to a network-based computing system to recognize voice commands when the user provides a specific voice command to stream to the network-based computing system and/or when the user provides a voice command that is not recognizable by the local system.

Hybrid voice command technique utilizing both on-device and cloud resources

A voice-activated system edge device cooperating with a remote command processor has a state machine defined by a listening mode state and a conversation monitoring mode state. The state machine transitions from the listening mode state to the conversation monitoring mode state in response to a wake word detection. A command accompanying the wake word is transmitted to the remote command processor for execution thereon. The conversation monitoring mode state is maintained for a conversation monitoring window time duration to receive a connection word accompanied by another command transmitted to the remote command processor for further execution thereon.

Wake word method to prolong the conversational state between human and a machine in edge devices

Systems and methods presented herein generally include multi-wake phrase detection executed on a single device utilizing multiple voice assistants. Systems and methods presented herein can further include continuously running a Voice Activity Detection (VAD) process which detects presence of human speech. The multi-wake phrase detection can activate when the VAD process detects human speech. Once activated, the multi-wake phrase detection can determine which (if any) of the wake phrases of the multiple voice assistants might be in the detected speech. Operation of the multi-wake phrase detection can have a low miss-rate. In some examples, operation of the multi-wake phrase detection can be granular to accomplish the low miss-rates at low power with a tolerance for false positives on wake phrase detection.

Low-power multi-voice assistants voice activation

A system can include a circuit holistically customized for the detection of commands in an audio or video input signal to meet certain application-specific requirements. The circuit can have a neural network topology that is hardwired to perform detection based on application-specific detection criteria. The hardwired custom circuit can provide improved energy efficiencies compared to similar functionality carried out using software and generic hardware modules. The system can also include a sound change trigger module and perform non-voiced sound detection.

Energy efficient custom deep learning circuits for always-on embedded applications

A speech encoder that analyzes and classifies each frame of speech as being periodic-like speech or non-periodic like speech where the speech encoder performs a different gain quantization process depending if the speech is periodic or not. If the speech is periodic, the improved speech encoder obtains the pitch gains from the unquantized weighted speech signal and performs a pre-vector quantization of the adaptive codebook gain Gfor each subframe of the frame before subframe processing begins and a closed-loop delayed decision vector quantization of the fixed codebook gain G. If the frame of speech is non-periodic, the speech encoder may use any known method of gain quantization. The result of quantizing gains of periodic speech in this manner results in a reduction of the number of bits required to represent the quantized gain information and for periodic speech, the ability to use the quantized pitch gain for the current subframe to search the fixed codebook for the fixed codebook excitation vector for the current subframe. Alternatively, the new gain quantization process which was used only for periodic signals may be extended to non-periodic signals as well. This second strategy results in a slightly higher bit rate than that for periodic signals that use the new gain quantization strategy, but is still lower than the prior art's bit rate. Yet another alternative is to use the new gain quantization process for all speech signals without distinguishing between periodic and non-periodic signals.

Speech gain quantization strategy

In accordance with one aspect of the invention, a selector supports the selection of a first encoding scheme or the second encoding scheme based upon the detection or absence of the triggering characteristic in the interval of the input speech signal. The first encoding scheme has a pitch pre-processing procedure for processing the input speech signal to form a revised speech signal biased toward an ideal voiced and stationary characteristic. The pre-processing procedure allows the encoder to fully capture the benefits of a bandwidth-efficient, long-term predictive procedure for a greater amount of speech components of an input speech signal than would otherwise be possible. In accordance with another aspect of the invention, the second encoding scheme entails a long-term prediction mode for encoding the pitch on a sub-frame by sub-frame basis. The long-term prediction mode is tailored to where the generally periodic component of the speech is generally not stationary or less than completely periodic and requires greater frequency of updates from the adaptive codebook to achieve a desired perceptual quality of the reproduced speech under a long-term predictive procedure.

Selection of scalar quantixation (SQ) and vector quantization (VQ) for speech coding

There is provided a method for use by a speech encoder to encode an input speech signal. The method comprises receiving the input speech signal; determining whether the input speech signal includes an active speech signal or an inactive speech signal; low-pass filtering the inactive speech signal to generate a narrowband inactive speech signal; high-pass filtering the inactive speech signal to generate a high-band inactive speech signal; encoding the narrowband inactive speech signal using a narrowband inactive speech encoder to generate an encoded narrowband inactive speech; generating a low-to-high auxiliary signal by the narrowband inactive speech encoder based on the narrowband inactive speech signal; encoding the high-band inactive speech signal using a wideband inactive speech encoder to generate an encoded wideband inactive speech based on the low-to-high auxiliary signal from the narrowband inactive speech encoder; and transmitting the encoded narrowband inactive speech and the encoded wideband inactive speech.

Decoder with embedded silence and background noise compression

Embedded silence and background noise compression

There is provided a voice activity detection method for indicating an active voice mode and an inactive voice mode. The method comprises receiving a first portion of an input signal; determining that the first portion of the input signal includes an active voice signal; indicating the active voice mode in response to the determining that the first portion of the input signal includes the active voice signal; receiving a second portion of the input signal immediately following the first portion of the input signal; determining that the second portion of the input signal includes an inactive voice signal; extending the indicating the active voice mode for a period of time after determining that the second portion of the input signal includes the inactive voice signal, wherein the period of time varies based on one or more conditions; and indicating the inactive voice mode after expiration of the period of time.

Adaptive voice mode extension for a voice activity detector

There are provided methods and systems for automated testing of echo cancellers using natural speech excitations and evaluating an echo canceller by transmitting a first signal to the echo canceller, wherein the first signal includes a first speech signal and a first marker signal, and wherein the first marker signal is transmitted a first period of time after the first speech signal is transmitted; receiving a second signal from the echo canceller, wherein the second signal includes a second speech signal and a second marker signal; aligning the first speech signal and the second speech signal using the first marker signal and the second marker signal; determining a choppiness of the second speech signal, when a non-linear processor of the echo canceller is on; and determining an audible echo, when a non-linear processor of the echo canceller is off.

Automated tools for testing echo cancellers using natural speech excitations

A method of adjusting an echo canceller comprises obtaining a first cross-correlation between a far-end signal and an error signal, wherein the error signal is generated by subtracting an output signal of an adaptive filter from a local-end signal; determining whether the first cross-correlation is above a pre-determined threshold; relocating the adaptive filter by a few samples if the determining determines that the first cross-correlation is above a pre-determined threshold; calculating a first improvement indicator parameter, wherein the first improvement indicator parameter is calculated after the relocating the adaptive filter by the few samples; determining whether the first improvement indicator parameter indicates a performance improvement by the adaptive filter after the relocating the adaptive filter by the few samples; calculating a gain based on the local-end signal and the error signal if the determining does not determine the performance improvement; and multiplying the adaptive filter by the gain.

Fast echo canceller reconvergence after TDM slips and echo level changes

A method of masking a residual echo signal by an echo canceller is provided. The method comprises receiving a far-end signal, adjusting filter coefficients of an adaptive filter in response to the far-end signal, generating an echo model signal based on the far-end signal using the adaptive filter, receiving a near-end signal, subtracting the echo model signal from the near-end signal to generate an output signal, defining a spectral mask based on the near-end signal, wherein the spectral mask is indicative of near-end spectral peaks and near-end spectral valleys, de-emphasizing the output signal in spectral regions of the near-end spectral peaks, and emphasizing the output signal in spectral regions of the near-end spectral valleys, wherein the de-emphasizing occurs during filter coefficients determination for the adaptive filter. A weighted filter may perform the de-emphasizing and the emphasizing operations, where the weighted filter uses medium term spectral characteristics of the near-end signal.

Growing community of inventors

Irvine, CA, United States of America

Adil Benyassine