# US 2015/0088497 A1 · Speech processing apparatus, speech processing method, and speech processing program

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2015/0088497 A1
- Document ID: US20150088497A1
- Status: Published application
- Kind: A1
- Application number: 14/495,012
- Filed: 2014-09-24
- Published: 2015-03-26
- Earliest claimed: 2013-09-26
- Assignees: Honda Motor Co., Ltd.
- Applicants: Honda Motor Co., Ltd.
- Inventors: Randy GOMEZ; Kazuhiro NAKADAI; Keisuke NAKAMURA
- Web page: https://patentgrep.com/patent/US20150088497A1

## CPC

- G10L 21/0208: Noise filtering
- G10L 15/20: Speech recognition techniques specially adapted for robustness in adverse environments, e.g. in noise, of stress induced speech (G10L21/02 takes precedence)
- G10L 2021/02082: {the noise being echo, reverberation of the speech}
- G10L 2021/02166: {Microphone arrays; Beamforming}
- G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING

## Abstract

A speech processing apparatus includes a sound collecting unit configured to collect sound signals, a sound source direction estimating unit configured to estimate a direction of a sound source of each sound signal collected by the sound collecting unit, a reverberation reducing filter calculating unit configured to calculate a reverberation reducing filter to be applied to the sound signals collected by the sound collecting unit, and a reduction processing unit configured to apply the reverberation reducing filter calculated by the reverberation reducing filter calculating unit to the sound signals, and the reverberation reducing filter calculating unit calculates the reverberation reducing filter to be applied based on the directions of the sound sources estimated by the sound source direction estimating unit.

## Claims

### Claim 1

A speech processing apparatus comprising:
a sound collecting unit configured to collect sound signals;
a sound source direction estimating unit configured to estimate a direction of a sound source of each sound signal collected by the sound collecting unit;
a reverberation reducing filter calculating unit configured to calculate a reverberation reducing filter to be applied to the sound signals collected by the sound collecting unit; and
a reduction processing unit configured to apply the reverberation reducing filter calculated by the reverberation reducing filter calculating unit to the sound signals,
wherein the reverberation reducing filter calculating unit calculates the reverberation reducing filter to be applied based on the directions of the sound sources estimated by the sound source direction estimating unit.

### Claim 2 (depends on 1)

The speech processing apparatus according to claim 1, wherein the reverberation reducing filter calculating unit calculates the reverberation reducing filter using an extension filter which is generated using a late reflection component of the sound signal and a response of the late reflection component of the direction of each of the sound sources.

### Claim 3 (depends on 1)

The speech processing apparatus according to claim 1, wherein the sound source direction estimating unit estimates the direction of the sound source using a feature vector of the single sound signal collected by the sound collecting unit and a probability model of the direction of each of the sound sources.

### Claim 4 (depends on 1)

The speech processing apparatus according to claim 1, further comprising a sound source separating unit configured to separate a full reverberant signal and a late reflection component from the sound signals collected by the sound collecting unit,
wherein the reverberation reducing filter calculating unit calculates the reverberation reducing filter using an extension filter which is generated using the late reflection component separated by the sound source separating unit and a response of the late reflection component of the direction of each of the sound sources.

### Claim 5 (depends on 4)

The speech processing apparatus according to claim 4, wherein the reduction processing unit reduces the late reflection component from the full reverberant signal separated by the sound source separating unit by applying the reverberation reducing filter calculated by the reverberation reducing filter calculating unit to the full reverberant signal.

### Claim 6 (depends on 4)

The speech processing apparatus according to claim 4, further comprising:
a first sound signal processing unit configured to calculate a first feature vector of the sound signals collected by the sound collecting unit based on a first room transfer function; and
a second sound signal processing unit configured to calculate a second feature vector of the sound signals collected by the sound collecting unit based on a second room transfer function,
wherein the sound source separating unit includes
a first sound source separating unit configured to separate the full reverberant signal based on the first feature vector calculated by the first sound signal processing unit, and
a second sound source separating unit configured to separate the late reflection component based on the second feature vector calculated by the second sound signal processing unit, and
wherein the reduction processing unit reduces the late reflection component separated by the second sound source separating unit from the full reverberant signal separated by the first sound source separating unit by applying the reverberation reducing filter calculated by the reverberation reducing filter calculating unit to the full reverberant signal.

### Claim 7 (depends on 1)

The speech processing apparatus according to claim 1, wherein the sound source direction estimating unit estimates the directions of the sound sources based on at least one of an image captured by an imaging unit and detection results of azimuth sensors attached to the vicinities of the sound sources.

### Claim 8

A speech processing method comprising:
a sound collecting step of collecting sound signals;
a sound source direction estimating step of estimating a direction of a sound source of each sound signal collected in the sound collecting step;
a reverberation reducing filter calculating step of calculating a reverberation reducing filter to be applied to the sound signals collected in the sound collecting step based on the directions of the sound sources estimated in the sound source direction estimating step; and
a reduction step of applying the reverberation reducing filter calculated in the reverberation reducing filter calculating step to the sound signals.

### Claim 9

A non-transitory computer-readable recording medium having recorded thereon a speech processing program causing a computer of a speech processing apparatus to perform:
a sound collecting procedure of collecting sound signals;
a sound source direction estimating procedure of estimating a direction of a sound source of each sound signal collected in the sound collecting procedure;
a reverberation reducing filter calculating procedure of calculating a reverberation reducing filter to be applied to the sound signals collected in the sound collecting procedure based on the directions of the sound sources estimated in the sound source direction estimating procedure; and
a reduction procedure of applying the reverberation reducing filter calculated in the reverberation reducing filter calculating procedure to the sound signals.
