Sound processing device and sound processing method

Patent metadata

Field	Value
Publication number	US-9542937-B2
Application number	US-201414148813-A
Country	US
Kind code	B2
Filing date	Jan 7, 2014
Priority date	Jan 15, 2013
Publication date	Jan 10, 2017
Grant date	Jan 10, 2017

Abstract

A sound processing device includes a noise suppression unit configured to suppress a noise component included in an input sound signal, an auxiliary noise addition unit configured to add auxiliary noise to the input sound signal, whose noise component has been suppressed by the noise suppression unit, to generate an auxiliary noise-added signal, a distortion calculation unit configured to calculate a degree of distortion of the auxiliary noise-added signal, and a control unit configured to control an addition amount by which the auxiliary noise addition unit adds the auxiliary noise based on the degree of distortion calculated by the distortion calculation unit.

First claim

What is claimed is:

1. A sound processing device, comprising:
at least one processor; and
at least one memory including computer program code,
wherein the at least one memory and the computer program code are configured to, with the at least on processor, cause the sound processing device to
suppress a noise component included in an input sound signal,
add auxiliary noise to the input sound signal, whose noise component has been suppressed by the noise suppression unit, to generate an auxiliary noise-added signal,
calculate a degree of distortion of the auxiliary noise-added signal,
estimate a speech recognition rate corresponding to the degree of distortion,
calculate a kurtosis ratio which is a ratio of a kurtosis of the auxiliary noise-added signal to a kurtosis of the input sound signal as the degree of distortion,
determine an addition amount based on the kurtosis ratio as an index value indicating the degree of distortion,
calculate a power spectrum based on the noise component,
calculate a complex noise-removed spectrum by subtracting the noise power from the power spectrum,
transform the complex noise-removed spectrum into the input sound signal whose noise component has been suppressed,
calculate a differential addition amount which is a difference between the determined addition amount and an ideal addition amount in which the speech recognition rate is the highest,
control the addition amount, by which the auxiliary noise is added to the input sound signal whose noise component has been suppressed to maximize a speech recognition rate based on the kurtosis ratio, by using the differential addition amount, and
perform a speech recognition process on the auxiliary noise-added signal.

2. The sound processing device according to claim 1 , wherein the at least one memory and the computer program code are further configured to, with the at least on processor, cause the sound processing device to estimate the speech recognition rate based on the degree of distortion of the auxiliary noise-added signal generated by suppressing the noise component with at least two types of suppression amounts, to select the suppression amount with which the estimated speech recognition rate is maximized, and to cause the noise suppression unit to suppress the noise component with the selected suppression amount.

3. The sound processing device according to claim 2 , wherein the at least one memory and the computer program code are further configured to, with the at least on processor, cause the sound processing device to control the addition amount of the auxiliary noise so as to maximize the estimated speech recognition rate with the selected suppression amount.

4. The sound processing device according to claim 1 , wherein the at least one memory and the computer program code are further configured to, with the at least on processor, cause the sound processing device to
calculate the degree of distortion for each component of the auxiliary noise-added signal, and
perform the speech recognition process so that the larger the degree of distortion of a component becomes, the smaller an influence of the component becomes.

5. A sound processing method, comprising:
detecting a noise component included in an input sound signal and suppressing the noise component detected from the input sound signal;
adding auxiliary noise to the input sound signal, whose noise component has been suppressed in the noise suppression step, to generate an auxiliary noise-added signal;
calculating a degree of distortion of the auxiliary noise-added signal;
estimating a speech recognition rate corresponding to the degree of distortion;
calculating a kurtosis ratio which is a ratio of a kurtosis of the auxiliary noise-added signal to a kurtosis of the input sound signal as the degree of distortion;
determining an addition amount based on the kurtosis ratio as an index value indicating the degree of distortion;
calculating a power spectrum based on the noise component;
calculating a complex noise-removed spectrum by subtracting the noise power from the power spectrum;
transforming the complex noise-removed spectrum into the input sound signal whose noise component has been suppressed;
calculating a differential addition amount which is a difference between the determined addition amount and an ideal addition amount in which the speech recognition rate is the highest;
controlling the addition amount, by which the auxiliary noise is added to the input sound signal whose noise component has been suppressed to maximize a speech recognition rate based on the kurtosis ratio, by using the differential addition amount; and
performing a speech recognition process on the auxiliary noise-added signal.

Assignees

• Honda Motor Co Ltd

Inventors

• Nakadai Kazuhiro
• Nakamura Keisuke
• Kimoto Daisuke

Classifications

• G10L15/20 Primary

  Speech recognition techniques specially adapted for robustness in adverse environments, e.g. in noise, of stress induced speech (G10L21/02 takes precedence) · CPC title

• G10L21/0208

  Noise filtering · CPC title