# US 2008/0270131 A1 · Method, preprocessor, speech recognition system, and program product for extracting target speech by removing noise

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

- Number: US 2008/0270131 A1
- Document ID: US20080270131A1
- Status: Published application
- Kind: A1
- Application number: 12/105,621
- Filed: 2008-04-18
- Published: 2008-10-30
- Earliest claimed: 2007-04-27
- Assignees: Nuance Communications, Inc.
- Inventors: Takashi Fukuda; Osamu Ichikawa; Masafumi Nishimura
- Web page: https://patentgrep.com/patent/US20080270131A1

## CPC

- 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 21/02: Speech enhancement, e.g. noise reduction or echo cancellation (reducing echo effects in line transmission systems H04B3/20; echo suppression in hands-free telephones H04M9/08)
- G10L 21/0272: Voice signal separating
- G10L 15/02: Feature extraction for speech recognition; Selection of recognition unit
- G10L 2021/02161: {Number of inputs available containing the signal or the noise to be suppressed}
- G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING

## Abstract

The present invention relates to a method, preprocessor, speech recognition system, and program product for extracting a target speech by removing noise. In an embodiment of the invention target speech is extracted from two input speeches, which are obtained through at least two speech input devices installed in different places in a space, applies a spectrum subtraction process by using a noise power spectrum (Uω) estimated by one or both of the two speech input devices (Xω(T)) and an arbitrary subtraction constant (α) to obtain a resultant subtracted power spectrum (Yω(T)). The invention further applies a gain control based on the two speech input devices to the resultant subtracted power spectrum to obtain a gain-controlled power spectrum (Dω(T)). The invention further applies a flooring process to said resultant gain-controlled power spectrum on the basis of arbitrary Flooring factor (β) to obtain a power spectrum for speech recognition (Zω(T)).

## Claims

### Claim 1

A method for extracting a target speech from two input speeches, which are obtained through at least two speech input devices installed in different places in a space, comprising:
Applying a spectrum subtraction process by using a noise power spectrum (Uω) estimated by one or both of said two speech input devices (Xω(T)) and an arbitrary subtraction constant (α) to obtain a resultant subtracted power spectrum (Yωo(T));
Applying a gain control based on the two speech input devices to said resultant subtracted power spectrum to obtain a gain-controlled power spectrum (Dω(T)); and;
Applying a flooring process to said resultant gain-controlled power spectrum on the basis of arbitrary Flooring factor (β) to obtain a power spectrum for speech recognition (Zω(T)).

### Claim 2 (depends on 1)

The method of claim 1, wherein said gain control uses a cross-power spectrum phase coefficient utilizing a cross-correlation of said two speech input devices.

### Claim 3 (depends on 1)

The method of claim 1, further comprising:
Performing an array process on said two speech input devices operable to enhance a speech from a target speaker before obtaining said subtracted power spectrum.

### Claim 4 (depends on 3)

The method of claim 3, wherein said array process is a delay-sum array process using said two speech input devices.

### Claim 5 (depends on 3)

The method of claim 3, wherein said array process is an adaptive array process using said two speech input devices.

### Claim 6 (depends on 1)

The method of claim 1, wherein said space is an inside of a car and said two speech input devices obtained through said two speech input devices operable to include a audio sound and/or a speech spoken by an interfering speaker as noise in said car.

### Claim 7 (depends on 2)

The method of claim 2, wherein said two speech input devices are one or more non-directional microphones.

### Claim 8 (depends on 2)

The method of claim 2, wherein said two speech input devices are one or more directional microphones.

### Claim 9 (depends on 1)

The method of claim 1, further comprising:
Forming a directivity in said car on the basis of a positional relationship between the two speech input devices installed in different places in said space and said target speaker.

### Claim 10 (depends on 1)

The method of claim 1, further comprising:
Outputting a speech recognition result in response to said power spectrum for speech recognition.

### Claim 11

A speech recognition system comprising:
A preprocessor for performing preprocessing before recognition of a target speech based on two input speeches obtained from at least two speech input devices installed in different places in a space.
A spectrum subtraction process section which is connected to said two speech input devices operable to apply a spectrum subtraction process using a noise power spectrum (Uω) estimated by one or both of said two input speeches devices (Xω(T)) and an arbitrary subtraction constant (α) to obtain a subtracted power spectrum (Yω(T));
A gain control section which is connected to said spectrum subtraction process section operable to apply gain control based on said two input speeches to a resultant subtracted power spectrum to obtain a gain-controlled power spectrum (Dω(T)); and
A flooring process section connected to said gain control section operable to apply a flooring process to said resultant gain-controlled power spectrum on a basis of a arbitrary Flooring factor (β) to obtain a power spectrum for speech recognition (Zω(T)).

### Claim 12 (depends on 11)

The system of claim 11, further comprising:
A recognizer connected to said preprocessor, wherein said speech recognition is outputted by entering said power spectrum for speech recognition.

### Claim 13

A computer program product comprising a computer useable medium including a computer readable program, wherein said computer readable program when executed on a computer causes the computer to perform the method steps to extracting a target speech from two input speeches, which are obtained through at least two speech input devices installed in different places in a space. The method comprising the steps of:
Applying a spectrum subtraction process by using a noise power spectrum (Uω) estimated by one or both of said two input speeches (Xω(T)) and an arbitrary subtraction constant (α) to obtain a subtracted power spectrum (Yω(T));
Applying a gain control based on said two input speeches to said subtracted power spectrum to obtain a gain-controlled power spectrum (Dω(T)); and
Applying a flooring process to said gain-controlled power spectrum on the basis of arbitrary flooring factor (β) to obtain a power spectrum for speech recognition (Zω(T)).

### Claim 14 (depends on 13)

The computer program product of claim 13, wherein said gain control uses a cross-power spectrum phase (CSP) coefficient utilizing a cross-correlation of said two input speeches.

### Claim 15 (depends on 13)

The computer program product of claim 13, further comprising:
Outputting a speech recognition result in response to said power spectrum for speech recognition as a preprocessing for extracting a target speech.
