Method and device for context-based voice recognition

Patent metadata

Field	Value
Publication number	US-2016019887-A1
Application number	US-201514616501-A
Country	US
Kind code	A1
Filing date	Feb 6, 2015
Priority date	Jul 21, 2014
Publication date	Jan 21, 2016
Grant date	—

Abstract

A method and a device of voice recognition are provided. The method involves receiving a voice signal, identifying a first voice recognition model in which context information associated with a situation at reception of the voice signal is not reflected and a second voice recognition model in which the context information is reflected, determining a weighted value of the first voice recognition model and a weighted value of the second voice recognition model, and recognizing a word in the voice signal by applying the determined weighted values to the first voice recognition model and the second voice recognition model.

First claim

What is claimed is:

1 . A method of voice recognition, the method comprising:
receiving a voice signal; identifying a first voice recognition model in which context information associated with a situation at reception of the voice signal is not reflected and a second voice recognition model in which the context information is reflected; determining a weighted value of the first voice recognition model and a weighted value of the second voice recognition model; and recognizing a word in the voice signal by applying the determined weighted values to the first voice recognition model and the second voice recognition model.

2 . The method of claim 1 , wherein the identifying of the second voice recognition model is performed based on the context information comprising at least one of environmental information and user profile information at reception of the voice signal.

3 . The method of claim 2 , wherein the environmental information comprises at least one of time at reception of the voice signal, weather at reception of the voice signal, location of a voice recognition device receiving the voice signal, and a moving speed of a voice recognition device receiving the voice signal.

4 . The method of claim 2 , wherein the user profile information comprises at least one of a gender, an age, a hometown, a hobby, and a marital status of a user.

5 . The method of claim 1 , wherein the first voice recognition model and the second voice recognition model comprise respective acoustic models and language models, and
the recognizing of the word in the voice signal comprises: determining at least one syllable comprised in the voice signal by using a phoneme probability based on an acoustic model of the first voice recognition model and a phoneme probability based on an acoustic model of the second voice recognition model; and recognizing a word comprising the at least one syllable by using a word probability based on a language model of the first voice recognition model and a word probability based on a language model of the second voice recognition model.

6 . The method of claim 5 , wherein the recognizing of the word in the voice signal further comprises:
recognizing a word subsequent to the recognized word by using a word probability based on the language model of the first voice recognition model and a word probability based on the language model of the second voice recognition model.

7 . The method of claim 1 , wherein the recognizing of the word in the voice signal is performed by applying, to an n-gram language model, the first voice recognition model and the second voice recognition model for which the weighted values are determined.

8 . The method of claim 1 , wherein the determining of the weighted values is performed based on a weighted value applied to a word recognized prior to a word being recognized.

9 . A method of voice recognition, the method comprising:
receiving a voice signal; verifying context information comprising at least one of environmental information and user profile information at reception of the voice signal; determining a weighted value of a first voice recognition model in which the context information is not reflected and a weighted value of a second voice recognition model in which the context information is reflected based on the context information; and recognizing a word in the voice signal by applying the determined weighted values to the first voice recognition model and the second voice recognition model and applying, to an n-gram language model, the first voice recognition model and the second voice recognition model to which the determined weighted values are applied.

10 . The method of claim 9 , wherein the first voice recognition model and the second voice recognition model comprise respective acoustic models and language models, and
the recognizing of the word in the voice signal comprises: determining at least one syllable comprised in the voice signal by using a phoneme probability based on an acoustic model of the first voice recognition model and a phoneme probability based on an acoustic model of the second voice recognition model; and recognizing a word comprising the at least one syllable by using a word probability based on a language model of the first voice recognition model and a word probability based on a language model of the second voice recognition model.

11 . The method of claim 10 , wherein the recognizing of the word in the voice signal further comprises:
recognizing a word subsequent to the recognized word by using a word probability based on the language model of the first voice recognition model and a word probability based on the language model of the second voice recognition model.

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

13 . A voice recognition device, the device comprising:
a receiver configured to receive a voice signal; a voice recognition model identifier configured to identify a first voice recognition model in which context information associated with a situation at reception of the voice signal is not reflected and a second voice recognition model in which the context information is reflected; a weighed value determiner configured to determine a weighted value of the first voice recognition model and a weighted value of the second voice recognition model; and a word recognizer configured to recognize a word in the voice signal by applying the determined weighted values to the first voice recognition model and the second voice recognition model.

14 . The device of claim 13 , wherein the voice recognition model identifier is configured to identify the second voice recognition model based on the context information comprising at least one of environmental information and user profile information at reception of the voice signal.

15 . The device of claim 14 , wherein the environmental information comprises at least one of time at reception of the voice signal, weather at reception of the voice signal, location of the voice recognition device receiving the voice signal, and a moving speed of the voice recognition device receiving the voice signal.

16 . The device of claim 14 , wherein the user profile information comprises at least one of a gender, an age, a hometown, a hobby, and a marital status of a user.

17 . The device of claim 13 , wherein the first voice recognition model and the second voice recognition model comprise respective acoustic models and language models, and
the word recognizer is configured to determine at least one syllable comprised in the voice signal by using a phoneme probability based on an acoustic model of the first voice recognition model and a phoneme probability based on an acoustic model of the second voice recognition model, and recognize a word comprising the at least one syllable by using a word probability based on a language model of the first voice recognition model and a word probability based on a language model of the second voice recognition model.

18 . The device of claim 17 , wherein the word recognizer is configured to further recognize a word subsequent to the recognized word by using a word probability based on the language model of the first voice recognition model and a word probability based on the language model of the second voice recognition model.

19 . The device of claim 13 , wherein the word recognizer is configured to recognize the word in the voice sig

Assignees

• Samsung Electronics Co Ltd

Inventors

• Kim Hyun-Jun
• Choi Young Sang

Classifications

• G10L15/08 Primary

  Speech classification or search · CPC title

• G10L15/197

  Probabilistic grammars, e.g. word n-grams · CPC title

• G10L2015/228

  of application context · CPC title

• G10L15/183

  using context dependencies, e.g. language models · CPC title