Channel selection apparatus, channel selection method, and program

Patent metadata

Field	Value
Publication number	US-12444403-B2
Application number	US-201917274394-A
Country	US
Kind code	B2
Filing date	Aug 28, 2019
Priority date	Sep 11, 2018
Publication date	Oct 14, 2025
Grant date	Oct 14, 2025

Abstract

A channel in which an utterance of a keyword is included is selected from acoustic signals of multiple channels. An addition unit 11 adds all channels of input voice signals of multiple channels to generate a composite voice signal of one channel. A keyword detection unit 12 generates a keyword detection result indicating a result of detecting an utterance of a predetermined keyword from a composite voice signal. A power calculation unit 13 calculates powers of channels based on input voice signals. A delay unit 14 delays the powers of the channels. When the keyword detection result indicates that the keyword was detected, a maximum power detection unit 15 selects, as an output channel, a channel having the maximum power among the powers of the channels of the input voice signals. A channel selection unit 16 selects the voice signal of the output channel from the input voice signals and outputs the selected voice signal.

First claim

What is claimed is:

1. A channel selection apparatus comprising:
input circuitry configured to receive input voice signals of a voice captured by a plurality of microphones over a plurality of channels of input voice signals, at least two of the input voice signals including a predetermined keyword;
addition circuitry configured to add all the received input voice signals of the voice of the plurality of channels to generate a composite voice signal of one channel;
keyword detection circuitry configured to generate a keyword detection result indicating a result of detecting an utterance of the predetermined keyword from the composite voice signal;
power calculation circuitry configured to calculate first power of an input voice signal in each channel of the plurality of channels;
maximum power detection circuitry configured to, when the keyword detection result indicates that the predetermined keyword has been detected, select a channel having the maximum power among respective powers of the input voice signals of the voice over the plurality of channels of the input voice signals as an output channel;
second power calculation circuitry configured to calculate second power of an input voice signal in said each channel of the plurality of channels in a time segment obtained by tracing back a predetermined amount of time from the input voice signals;
weight calculation circuitry configured to calculate a weight having a value that increases the larger the first power output by the power calculation circuitry is than the second power output by the second power calculation circuitry,
wherein the maximum power detection circuitry is further configured to detect, as an output channel, the channel having the maximum power among the respective powers of the input voice signals of the voice over the plurality of channels, and the channel is obtained by weighting the powers of the input voice signals of the voice over the plurality of channels output by the power calculation circuitry with the weight; and

speech recognition circuitry configured to subject the output channel to speech recognition.

2. A non-transitory computer-readable recording medium on which a program recorded thereon for causing a computer to function as the channel selection apparatus according to claim 1 .

3. A channel selection apparatus comprising:
input circuitry configured to receive input voice signals of a voice captured by a plurality of microphones over M channels of input voice signals, at least two of the input voice signals including a predetermined keyword;
power calculation circuitry configured to calculate first power of an input voice signal in each channel of the M channels based on input voice signals of the M channels, where M is an integer that is 3 or more;
candidate selection circuitry configured to select, as candidate channels, K channels with large powers among the channels of the input voice signals, where K is an integer that is 1 or more and less than M;
keyword detection circuitry configured to generate a keyword detection result indicating a result of detecting an utterance of the predetermined keyword based on the input voice signals of the candidate channels;
maximum power detection circuitry configured to select, as an output channel, a channel having the maximum power among respective powers of the input voice signals of the voice over the candidate channels for which the keyword detection result indicates that a keyword has been detected;
second power calculation circuitry configured to calculate second power of an input voice signal in said each channel of the plurality of channels in a time segment obtained by tracing back a predetermined amount of time from the input voice signals;
weight calculation circuitry configured to calculate a weight having a value that increases the larger the first power output by the power calculation circuitry is than the second power output by the second power calculation circuitry,
wherein the candidate selection circuitry configured to select, as the candidate channels, K channels with large powers obtained by weighting the powers of the channels output by the power calculation circuitry with the weight, and
the maximum power detection circuitry further configured to detect, as an output channel, the channel having the maximum power among the respective powers of the input voice signals of the voice over the plurality of channels, and the channel is obtained by weighting the powers of the input voice signals of the voice over the plurality of channels output by the power calculation circuitry with the weight; and

speech recognition circuitry configured to subject the output channel to speech recognition.

4. A non-transitory computer-readable recording medium on which a program recorded thereon for causing a computer to function as the channel selection apparatus according to claim 3 .

5. An apparatus for channel selection comprising a processor configured to execute operations comprising:
receiving input voice signals of a voice captured by a plurality of microphones over channels of input voice signals, at least two of the input voice signals including a predetermined keyword;
adding, by processing circuitry of a channel selection apparatus, all the received input voice signals of the voice over a plurality of channels to generate a composite voice signal of one channel;
generating, by the processing circuitry of the channel selection apparatus, a keyword detection result indicating a result of detecting an utterance of the predetermined keyword from the composite voice signal;
calculating, by the processing circuitry of the channel selection apparatus, first power of an input voice signal in each channel of the plurality of channels;
selecting, by the processing circuitry of the channel selection apparatus, as an output channel, a channel having the maximum power among respective powers of the input voice signals of the voice over the plurality of channels of input voice signals when the keyword detection result indicates that a keyword was detected to output the keyword through the selected output channel;
calculating second power of an input voice signal in said each channel of the plurality of channels in a time segment obtained by tracing back a predetermined amount of time from the input voice signals; and
calculating a weight having a value that increases the larger the first power is than the second power,
wherein the maximum power detection circuitry is further configured to detect, as an output channel, the channel having the maximum power among the respective powers of the input voice signals of the voice over the plurality of channels, and the channel is obtained by weighting the powers of the input voice signals of the voice over the plurality of channels; and

specifying the output channel to perform speech recognition.

Assignees

• Ntt Inc

Inventors

• Kobayashi Kazunori
• Saito Shoichiro
• Ito Hiroaki

Classifications

• G10L2015/088

  Word spotting · CPC title

• G10L25/78

  Detection of presence or absence of voice signals (switching of direction of transmission by voice frequency in two-way loud-speaking telephone systems H04M9/10) · CPC title

• G10L21/0272

  Voice signal separating · CPC title

• G10L15/083

  Recognition networks (G10L15/142, G10L15/16 take precedence) · CPC title

• Y02D30/70

  in wireless communication networks · CPC title