Texas Instruments Corporation

Amazon Technologies, Inc.

Mohamed Farouk Mansour

Srinath Hosur

June Chul Roh

Pierre Bertrand

Vijay Pothukuchi

Shobha Devi Kuruba Buchannagari

Corina Ioana Ionita

Robert Ayrapetian

Carlos Renato Nakagawa

Deric Wayne Waters

Carlo Murgia

Murtaza Ali

Philip Ryan Hilmes

Lars Jorgensen

Domingo G Garcia

Guangdong Pan

Parind Shah

Yuan-Yen Tai

June Chui Roh

Ariton E Xhafa

Amit Singh Chhetri

Sthanunathan Ramakrishnan

Trausti Thor Kristjansson

Ludger Solbach

Wontak Kim

Srivatsan Kandadai

Spencer Russell

Yuancheng Luo

Borham Lee

JuneChul Roh

Stefan Vlaski

Jun Chul Roh

A system configured to improve user localization used to determine a listening position and/or user orientation for a device map. Multiple devices may generate audio data representing user speech and the system may use the audio data to determine a first spatial likelihood function (SLF) based on angle measurements, determine a second SLF based on timing information, and determine a location of the user based on a combination of the two SLFs. The SLFs represent the environment using a grid comprising a plurality of grid cells, and each grid cell has a value indicating a likelihood that the grid cell corresponds to the location of the user. An individual device may generate a portion of the angle measurements based on multi-channel audio data generated using multiple microphones of the device, while the system may generate the timing information based on single-channel audio data received from each of the multiple devices.

User localization

Disclosed are techniques for an improved method for performing sound source localization (SSL) to determine a direction of arrival of an audible sound using a combination of timing information and amplitude information. For example, a device may decompose an observed sound field into directional components, then estimate a time-delay likelihood value and an energy-based likelihood value for each of the directional components. Using a combination of these likelihood values, the device can determine the direction of arrival corresponding to a maximum likelihood value. In some examples, the device may perform Acoustic Wave Decomposition processing to determine the directional components. In order to reduce a processing consumption associated with performing AWD processing, the device splits this process into two phases: a search phase that selects a subset of a device dictionary to reduce a complexity, and a decomposition phase that solves an optimization problem using the subset of the device dictionary.

Sound source localization using acoustic wave decomposition

A method for communicating over a wireless backhaul channel comprising generating a radio frame comprising a plurality of time slots, wherein each time slot comprises a plurality of symbols in time and a plurality of sub-carriers in a system bandwidth, broadcasting a broadcast channel signal comprising a transmission schedule to a plurality of remote units in a number of consecutive sub-carriers centered about a direct current (DC) sub-carrier in at least one of the time slots in the radio frame regardless of the system bandwidth, and transmitting a downlink (DL) control channel signal and a DL data channel signal to a first of the remote units, wherein the DL data channel signal is transmitted by employing a single carrier block transmission scheme comprising a Discrete Fourier Transform (DFT) spreading for frequency diversity.

NLOS wireless backhaul downlink communication

Disclosed are techniques for performing ray-based acoustic modeling that models scattering of acoustic waves by a surface of a device. The acoustic modeling uses two parameters, a room response representing acoustics and geometry of a room and a device response representing acoustics and geometry of the device. The room response is determined using ray-based acoustic modeling, such as ray tracing. The device response can be measured in an actual environment or simulated and represents an acoustic response of the device to individual acoustic plane waves. The device applies a superposition of the room response and the plane wave scattering from the device response to determine acoustic pressure values and generate microphone audio data. The device can estimate room impulse response (RIR) data using data from the microphones, and can use the RIR data to perform audio processing such as sound equalization, acoustic echo cancellation, audio beamforming, and/or the like.

Surface augmented ray-based acoustic modeling

Techniques for improving adaptive interference cancellation (AIC) using cascaded AIC algorithms are described. To improve an accuracy of detecting speech, a device may perform a first stage of AIC to generate isolated audio data and may generate speech mask data indicating time windows when speech is detected in the isolated audio data. Based on the speech mask data, the device may perform second AIC to generate output audio data, with adaptation of the adaptive filter enabled when the speech is not detected and disabled when the speech is detected. Thus, the first AIC improves the accuracy with which the device detects that speech is present and the second AIC reduces distortion in the output audio data by not updating filter coefficient values when the speech is present. The first AIC may use playback audio data, microphone audio data or beamformed audio data as reference signals.

Cascaded adaptive interference cancellation algorithms

A method for uplink (UL) wireless backhaul communication at a wireless backhaul remote unit in a radio access network comprising receiving a configuration for radio frames and a transmission schedule through a downlink (DL) physical layer broadcast channel, wherein the transmission schedule comprises a transmission allocation for the remote unit, generating a UL data frame, wherein generating the UL data frame comprises performing forward error correction (FEC) encoding on a data bit stream to generate a plurality of FEC codewords, wherein performing the FEC encoding comprises performing Reed Solomon (RS) encoding on the data bit stream to generate a plurality of RS codewords, performing byte interleaving on the RS codewords, and performing Turbo encoding on the byte interleaved RS codewords to generate one or more Turbo codewords, wherein each Turbo codeword is encoded from more than one RS codeword, and transmitting the UL data frame according to the transmission allocation.

NLOS wireless backhaul uplink communication

Disclosed are techniques for an improved method for performing Acoustic Wave Decomposition (AWD) processing that reduces a complexity and processing consumption. The improved method enables a device to perform AWD processing to decompose an observed sound field into directional components, enabling the device to perform additional processing such as sound source separation, dereverberation, sound source localization, sound field reconstruction, and/or the like. The improved method splits the solution to two phases: a search phase that selects a subset of a device dictionary to reduce a complexity, and a decomposition phase that solves an optimization problem using the subset of the device dictionary.

Multi-stage solver for acoustic wave decomposition

A device has a microphone array that acquires sound data and a camera that acquires image data. A portion of the device may be moveable by one or more actuators. Responsive to the user, the portion of the device is moved toward an estimated direction of the user. The estimated direction is based on sensor data including the sound data and the image data. First variance values for individual sound direction values are calculated. Data derived from the image data or data from other sensors may be used to modify the first variance values and determine second data comprising second variances. The second data may be processed to determine the estimated direction of the user. For example, the second data may be processed by both a forward and a backward Kalman filter, and the output combined to determine an estimated direction toward the user.

System to determine direction toward user

A system and method for providing error control coding for backhaul applications are disclosed. Data is first encoded using Reed-Solomon (RS) coding. The output RS blocks are then turbo coded. The size of the output RS blocks is selected to match the input of the turbo encoder. The bits from the RS blocks may be interleaved to create the input turbo blocks. Cyclic Redundancy Check (CRC) parity bits may be added to the data prior to RS coding.

Forward error control coding

A system that performs early reflections filtering to suppress early reflections and improve sound source localization (SSL). During music playback and/or when a device is placed in a corner, acoustic reflections from nearby surfaces get boosted due to constructive interference, negatively impacting SSL and other processing of the device. To suppress these early reflections, the device uses an Early Reflections Filter (ERF) that makes use of Linear Prediction Coding (LPC), which is already being performed during speech processing. For example, the device generates raw audio signals using multi-channel LPC coefficients and then uses single-channel LPC coefficients for each raw audio signal in order to generate a filter that estimates the reflections. The device then uses this filter to suppress the early reflections and generate filtered audio signals, thus resulting in better audio processing and better overall device performance.

Filtering early reflections

A system that performs sound source localization (SSL) using acoustic wave decomposition (AWD) or an approximation. When a device detects a wakeword represented in audio data, the device performs SSL processing in order to determine a position of the user relative to the device (e.g., estimate angle of the user). The device calculates noise statistics based on first audio data representing the wakeword and second audio data preceding the wakeword. Thus, upon detecting the wakeword, the device calculates the noise statistics and a signal quality metric corresponding to the wakeword. In addition, the device uses Multi-Channel Linear Prediction Coding (MCLPC) coefficients to average out the room impulse response. Using the noise statistics, the MCLPC coefficients, and the audio data, the device performs AWD processing to decompose the sound field to disjoint acoustic plane waves, enabling the device to identify the most likely direction for the line-of-sight component of speech.

Sound source localization using wave decomposition

Techniques for improving microphone noise suppression are provided. A system for noise-suppression may include a beam selector component that applies logic to select a beam most likely corresponding to a direction of a noise source and keeps the beam selection steady rather than switching the beam too often to avoid processing complications. The selected beam may be used as a reference in an adaptive filter which outputs a noise estimate. The noise estimate and raw microphone data may be used to adapt the adaptive filter. A parallel filter which adapts after a time delay may be applied to the reference in order to prevent interference. An attenuation factor may be used to scale the noise estimate based on noise diffuseness, signal quality, and/or a gain limit. The scaled noise estimate may be subtracted from microphone input data to produce output audio data with improved signal quality and maintained signal coherence.

Microphone noise suppression with beamforming

A system configured to improve wakeword detection. The system may selectively rectify (e.g., attenuate) a portion of an audio signal based on energy statistics corresponding to a keyword (e.g., wakeword). For example, a device may perform echo cancellation to generate isolated audio data, may use the energy statistics to calculate signal quality metric values for a plurality of frequency bands of the isolated audio data, and may select a fixed number of frequency bands (e.g., 5-10%) associated with lowest signal quality metric values. To detect a specific keyword, the system determines a threshold &#x3bb;(f) corresponding to an expected energy value at each frequency band. During runtime, the device determines signal quality metric values by subtracting residual music from the expected energy values. Thus, the device attenuates only a portion of the total number of frequency bands that include more energy than expected based on the energy statistics of the wakeword.

Growing community of inventors

Richardson, TX, United States of America

Mohamed Farouk Mansour