Speech recognition method and apparatus

Patent metadata

Field	Value
Publication number	US-10770065-B2
Application number	US-201715842430-A
Country	US
Kind code	B2
Filing date	Dec 14, 2017
Priority date	Dec 19, 2016
Publication date	Sep 8, 2020
Grant date	Sep 8, 2020

Abstract

A speech recognition method and a speech recognition apparatus which pre-download a speech recognition model predicted to be used and use the speech recognition model in speech recognition is provided. The speech recognition method, performed by the speech recognition apparatus, includes determining a speech recognition model, based on user information downloading the speech recognition model, performing speech recognition, based on the speech recognition model, and outputting a result of performing the speech recognition.

First claim

What is claimed is:

1. A speech recognition method, performed by a speech recognition apparatus, the speech recognition method comprising:
determining a speech recognition model, based on user information, the determining of the speech recognition model comprising determining a predictive language model included in the speech recognition model based on language models included in a range, the range being determined based on an utterance predicted to be uttered by a user of the speech recognition apparatus;
downloading the speech recognition model;
performing speech recognition, based on the speech recognition model; and
outputting a result of performing the speech recognition.

2. The speech recognition method of claim 1 , wherein the speech recognition model comprises at least one of a pre-processing model, an acoustic model, or a language model.

3. The speech recognition method of claim 1 , further comprising obtaining the user information,
wherein the user information comprises at least one of information about a characteristic of the user, information about an environment in which the user is located, information about a vehicle in which the user is located, information about a destination of the user, or information input by the user.

4. The speech recognition method of claim 1 ,
wherein the predictive language model is related to the predicted utterance, and
wherein the determining of the speech recognition model further comprises:
determining a domain of a language model, based on the user information;
determining a range of language models comprised in the domain, based on information about a characteristic of the user which is comprised in the user information; and
determining at least one language model to be the predictive language model, wherein the range comprises the at least one language model.

5. The speech recognition method of claim 1 , wherein the determining of the speech recognition model further comprises determining the speech recognition model by referring to a capacity of a memory in the speech recognition apparatus.

6. The speech recognition method of claim 1 , wherein the determining of the speech recognition model further comprises determining the speech recognition model according to at least one of when a user input is received from the user, when there is a change in an environment in which the user is located, or when a use period of the speech recognition model has expired.

7. The speech recognition method of claim 1 , wherein the downloading of the speech recognition model comprises downloading the speech recognition model from at least one of a server connected to the speech recognition apparatus or an electronic device connected to the speech recognition apparatus.

8. The speech recognition method of claim 1 ,
wherein the speech recognition model comprises the predictive language model related to the predicted utterance,
wherein the determining of the speech recognition model further comprises determining the predictive language model comprising a language model related to a first point comprised in a route, based on information about the route the user takes to move to a destination, and
wherein the downloading of the speech recognition model comprises downloading the language model related to the first point before arriving at the first point, based on predicting that a network status of the speech recognition apparatus for communication at the first point will be insufficient.

9. The speech recognition method of claim 1 , further comprising determining a usage method of the speech recognition model,
wherein the performing of the speech recognition comprises performing the speech recognition by using the speech recognition model, based on the usage method.

10. The speech recognition method of claim 1 , wherein a usage method of the speech recognition model comprises at least one of a use period of the speech recognition model, a method of processing the speech recognition model after the speech recognition model is used, or reliability of the speech recognition model.

11. The speech recognition method of claim 9 ,
wherein the performing of the speech recognition based on the usage method comprises performing the speech recognition by using the speech recognition model, during a use period of the speech recognition model which is comprised in the usage method, and
wherein the speech recognition method further comprises, after the use period of the speech recognition model is expired, deleting or deactivating the speech recognition model, based on the usage method.

12. The speech recognition method of claim 9 ,
wherein the speech recognition model comprises the predictive language model related to the predicted utterance,
wherein the determining of the speech recognition model further comprises determining the predictive language model related to a foreign language used in a region where the user is to be located, based on information about a schedule of the user which is stored in the speech recognition apparatus,
wherein the determining of the usage method of the speech recognition model comprises determining to activate the predictive language model in a period during which the user is to be located in the region, based on the schedule of the user, and
wherein the outputting of the result of performing the speech recognition comprises outputting a result of performing translation between the foreign language and a language used by the user, based on the predictive language model.

13. The speech recognition method of claim 1 ,
wherein a depth indicates top and bottom ranges of language models included in the predictive language model when the language models have a hierarchical structure, and
wherein a root indicates fields of the language models included in the predictive language model when the language models have a similarity therebetween.

14. A speech recognition apparatus comprising:
a receiver configured to receive speech of a user of the speech recognition apparatus;
at least one processor configured to:
determine a speech recognition model, based on user information, the determining of the speech recognition model comprising determining a predictive language model included in the speech recognition model based on language models included in a range, the range being determined based on an utterance predicted to be uttered by the user, and
perform speech recognition on the speech of the user, based on the speech recognition model;

a communicator configured to download the speech recognition model; and
an output interface configured to output a result of performing the speech recognition.

15. The speech recognition apparatus of claim 14 , wherein the speech recognition model comprises at least one of a pre-processing model, an acoustic model, or a language model.

16. The speech recognition apparatus of claim 14 ,
wherein the at least one processor is further configured to obtain the user information, and
wherein the user information comprises at least one of information about a characteristic of the user, information about an environment in which the user is located, information about a vehicle in which the user is located, information about a destination of the user, or information input by the user.

17. The speech recognition apparatus of claim 14 ,
wherein the speech recognition model comprises the predictive language model related to the predicted utterance, and
wherein, when the speech recognition model is determined, the at least one processor is further configured to:
determine a do

Assignees

• Samsung Electronics Co Ltd

Inventors

• Kim Sang-Yoon
• Kim Sung-Soo
• Kim Il-Hwan
• Lee Kyung-Min
• Kim Nam-Hoon
• Ryu Jong-Youb
• Lee Jae-Won

Classifications

• G10L19/04

  using predictive techniques · CPC title

• G10L17/02 Primary

  Preprocessing operations, e.g. segment selection; Pattern representation or modelling, e.g. based on linear discriminant analysis [LDA] or principal components; Feature selection or extraction · CPC title

• G10L15/183

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

• G10L15/26

  Speech to text systems (G10L15/08 takes precedence) · CPC title

• G10L15/22

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