Method and apparatus for speech recognition

Patent metadata

Field	Value
Publication number	US-9953647-B2
Application number	US-201614993439-A
Country	US
Kind code	B2
Filing date	Jan 12, 2016
Priority date	Jan 19, 2015
Publication date	Apr 24, 2018
Grant date	Apr 24, 2018

Abstract

A method and apparatus for speech recognition are provided. The method and the apparatus calculate signal to noise ratios (SNRs) of speech signals from a user received at speech recognition apparatuses. The method and the apparatus recognize a reference speech signal having a maximum SNR among the SNRs.

First claim

What is claimed is:

1. A processor implemented speech recognition method of a speech recognition apparatus, comprising:
capturing a speech signal from a user;
calculating a signal to noise ratio (SNR) of the captured speech signal;
selectively recognizing the captured speech signal in response to the SNR of the captured speech signal being greater than an SNR of the speech signal received at another speech recognition apparatus among plural speech recognition apparatuses,
wherein the selective recognizing of the captured speech signal includes:
selecting between the speech recognition apparatus and the other speech recognition apparatus to recognize the captured speech signal based on a determined amount of available resources of the speech recognition apparatus, and
transmitting, from the speech recognition apparatus, the captured speech signal or a feature value extracted from the captured speech signal to the other speech recognition apparatus in response to the determined amount of available resources being determined to fail to meet a threshold, so the recognizing of the captured speech signal is performed by the other speech recognition apparatus based on the captured speech signal or the extracted feature value.

2. The method of claim 1 , wherein the calculating of the SNRs comprises calculating the SNR by separating noise and a speech of the user comprised in the captured speech signal.

3. The method of claim 1 , wherein the speech recognition apparatus is selected to perform the recognizing of the captured speech signal in response to the amount of available resources being determined to meet the threshold.

4. The method of claim 1 , wherein the calculating of the SNR comprises, when a battery condition of the speech recognition apparatus fails to meet another threshold value, transmitting the captured speech signal to the other speech recognition apparatus so the calculating of the SNR of the captured speech is performed by the other speech recognition apparatus.

5. The method of claim 1 , wherein the SNR of the captured speech signal being greater than the SNR of the speech signal received at the other speech recognition apparatus is determined by the other speech recognition apparatus having received SNRs from the plural speech recognition apparatuses or a communication among the plural speech recognition apparatuses.

6. The method of claim 1 , further comprising:
determining whether the user issues a command to any speech recognition apparatus among the plural speech recognition apparatuses; and
receiving the speech signals through the plural speech recognition apparatuses in response to a determination that the user issues the command to any speech recognition apparatus among the plural speech recognition apparatuses.

7. The method of claim 1 , further comprising:
verifying a user command based on the recognition of the captured speech signal and transmitting the user command to one of the plural speech recognition apparatuses corresponding to the command,
wherein the one speech recognition apparatus corresponding to the command executes the command.

8. The method of claim 1 , wherein the plural speech recognition apparatuses are positioned at different positions with respect to the user.

9. The method of claim 1 , wherein the plural speech recognition apparatuses correspond to objects in an Internet of Things (IoT) environment.

10. The method of claim 1 , wherein the plural speech recognition apparatuses comprises at least one of a smartphone, a smart television, a smart watch, a smart washing machine, a smart camera, a smart audio, a smart bicycle, smart glasses, a remote controller, a smart refrigerator, a smart air conditioner, a laptop, a personal digital assistant (PDA), and a tablet.

11. A non-transitory computer-readable storage medium comprising a program comprising instructions to cause a processor to perform the method of claim 1 .

12. A speech recognition apparatus, comprising:
a microphone configured to capture a speech signal from a user; and
a processor configured to calculate a signal to noise ratio (SNR) of captured speech signal and selectively recognize the captured speech signal in response to the SNR of the captured speech signal being greater than an SNR of the speech signal received at another speech recognition apparatus among plural speech recognition apparatuses,
wherein the selective recognizing of the captured speech signal includes selecting between the speech recognition apparatus and the other speech recognition apparatus to recognize the captured speech signal based on a determined amount of available resources of the speech recognition apparatus, and
wherein the processor is configured to transmit the captured speech signal or a feature value extracted from the captured speech signal to the other speech recognition apparatus in response to the amount of available resources failing to meet a threshold so the recognizing of the captured speech signal is performed by the other speech recognition apparatus based on the captured speech signal or the extracted feature value.

13. The apparatus of claim 12 , wherein the processor is configured to calculate the SNR by separating noise from speech of the user comprised in the captured speech signal.

14. The apparatus of claim 12 , wherein, for the selective recognizing, the processor is configured to determine that the captured speech signal is to be recognized by the speech recognition apparatus when the determined amount of available resources of the speech recognition apparatus meets the threshold.

15. The apparatus of claim 12 , wherein the processor is configured to transmit the extracted feature value to the other speech recognition apparatus in response to the amount of available resources failing to meet the threshold.

16. The apparatus of claim 12 , wherein the processor is configured to determine whether the user issues a command to the speech recognition apparatus or to any of the plural speech recognition apparatuses, and the microphone is configured to capture the speech signal in response to a determination that the user issues the command.

17. The apparatus of claim 12 , wherein the speech recognition apparatus is configured to correspond to an object in an Internet of Things (IoT) environment.

18. The apparatus of claim 12 , wherein, in response to the SNR being a high speech signal, the processor performs speech recognition at a high recognition rate.

19. The apparatus of claim 14 , wherein, when the other speech recognition apparatus indicates a battery condition better than a battery condition of the speech recognition apparatus or the other speech recognition apparatus is indicated to be disposed closer to the user than the speech recognition apparatus, the determined amount of available resources of the speech recognition apparatus is determined to fail to meet the threshold.

Assignees

• Samsung Electronics Co Ltd

Inventors

• Mun Minyoung
• Choi Youngsang

Classifications

• G10L2015/223

  Execution procedure of a spoken command · CPC title

• G10L15/32

  Multiple recognisers used in sequence or in parallel; Score combination systems therefor, e.g. voting systems · CPC title

• G10L15/22 Primary

  Procedures used during a speech recognition process, e.g. man-machine dialogue · CPC title

• G10L21/0216

  characterised by the method used for estimating noise · CPC title

• G10L25/84

  for discriminating voice from noise · CPC title