US 9,818,425 B1Granted

Parallel output paths for acoustic echo cancellation

AssigneeAmazon Technologies, Inc.

InventorsRobert Ayrapetian, Philip Ryan Hilmes, Wai Chung Chu, Hyeong Cheol Kim, Yuwen Su

Life of this patent

Estimating the term…

Loading patent…

Abstract

An echo cancellation system that generates multiple output paths, enabling Automatic Speech Recognition (ASR) processing in parallel with voice communication. For single direction AEC (e.g., ASR processing), the system prioritizes speech from a single user and ignores other speech by selecting a single directional output from a plurality of directional outputs as a first output path. For multi-directional AEC (e.g., voice communication), the system includes all speech by combining the plurality of directional outputs as a second output path. The system may use a weighted sum technique, such that each directional output is represented in the combined output based on a corresponding signal metric, or an equal weighting technique, such that a first group of directional outputs having a higher signal metric may be equally weighted using a first weight while a second group of directional outputs having a lower signal metric may be equally weighted using a second weight.

Claims24

Dependent claims

A computer-implemented method for generating different audio outputs, the method comprising:

receiving, as part of a telecommunication connection, first voice data from a mobile phone, the first voice data corresponding to first speech;
sending the first voice data to a wireless speaker for playback;
receiving second voice data from a microphone array, the second voice data including a first representation of second speech;
determining, using the second voice data, first processed data, the first processed data corresponding to a first direction and including a second representation of the second speech;
determining, using the second voice data, a second processed data, the second processed data corresponding to a second direction and including a third representation of the second speech;
determining a first signal to noise ratio of the first processed data;
determining a second signal to noise ratio of the second processed data;
determining that the first signal to noise ratio is greater than the second signal to noise ratio;
generating, using the first processed data and the second processed data, first output data including a fourth representation of the speech;
sending the first processed data to a server for performing speech recognition to determine and execute a command;
and
sending the first output data to the mobile phone as part of the telecommunication connection.

9 more elements

A computer-implemented method, comprising:

capturing, using a microphone array, first voice audio data including a first representation of speech;
determining, using the first voice audio data, first processed audio data, the first processed audio data corresponding to a first direction and including a second representation of the speech;
determining, using the first voice audio data, second processed audio data, the second processed audio data corresponding to a second direction and including a third representation of the speech;
determining a first signal metric associated with a first frame of the first processed audio data;
determining a second signal metric associated with a first frame of the second processed audio data;
determining, based on the first signal metric, a first weight associated with the first frame of the first processed audio data;
determining, based on the second signal metric, a second weight associated with the first frame of the second processed audio data;
generating a first frame of first output audio data including a fourth representation of the speech, the first frame of the first output audio data including a first portion corresponding to the first frame of the first processed audio data based on the first weight and a second portion corresponding to the first frame of the second processed audio data based on the second weight;
sending the first frame of the first processed audio data to a first remote device;
and
sending the first frame of the first output audio data to a second remote device.

8 more elements

A device, comprising:

at least one processor;
and
a memory device including instructions operable to be executed by the at least one processor to configure the device to:
capture, using a microphone array, first voice audio data including a first representation of speech;
determine, using the first voice audio data, first processed audio data, the first processed audio data corresponding to a first direction and including a second representation of the speech;
determine, using the first voice audio data, second processed audio data, the second processed audio data corresponding to a second direction and including a third representation of the speech;
determine a first signal metric associated with a first frame of the first processed audio data;
determine a second signal metric associated with a first frame of the second processed audio data;
determine, based on the first signal metric, a first weight associated with the first frame of the first processed audio data;
determine, based on the second signal metric, a second weight associated with the first frame of the second processed audio data;
generate a first frame of first output audio data including a fourth representation of the speech, the first frame of the first output audio data including a first portion corresponding to the first frame of the first processed audio data based on the first weight and a second portion corresponding to the first frame of the second processed audio data based on the second weight;
send the first frame of the first processed audio data to a first remote device;
and
send the first frame of the first output audio data to a second remote device.

11 more elements

Dependencies are inferred from the claim text.

Facts

Application: 15/185,799Filed 17 Jun 2016
Granted: 14 Nov 2017
Term ends: Not estimatedThe source records don’t support an estimate.

Classification6 codes

G10L 21/0224
Characterised by the method used for estimating noise: processing in the time domain
G10L 15/30
Constructional details of speech recognition systems: distributed recognition, e.g. in client-server systems, for mobile phones or network…
G10L 2015/223
Procedures used during a speech recognition process, e.g. man-machine dialogue: execution procedure of a spoken command
G10L 2021/02082
Noise filtering: the noise being echo, reverberation of the speech