SPEECH RECOGNITION DEVICE AND THE OPERATION METHOD THEREOF

• US 20150081288A1
• Filed: 03/25/2014
• Published: 03/19/2015
• Est. Priority Date: 09/17/2013
• Status: Active Grant

• Current Assignee: Electronics and Telecommunications Research Institute
• Original Assignee: Electronics and Telecommunications Research Institute
• Inventors: KIM, Dong-Hyun
• Granted Patent: US 9,514,751 B2
• Time in Patent Office: Days
• US Class Current: 704/231
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models
• CPC Class Codes: G10L 15/187 Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams
• CPC Class Codes: G10L 15/30 Distributed recognition, e.g. in client-server systems, for mobile phones or network applications
• CPC Class Codes: G10L 2015/228 of application context

Abstract

Described herein is a speech recognition device comprising: a communication module receiving speech data corresponding to speech input from a speech recognition terminal and multi-sensor data corresponding to input environment of the speech; a model selection module selecting a language and acoustic model corresponding to the multi-sensor data among a plurality of language and acoustic models classified according to the speech input environment on the basis of previous multi-sensor data; and a speech recognition module controlling the communication module to apply a feature vector extracted from the speech data to the language and acoustic model and transmit speech recognition result for the speech data to the speech recognition terminal.

Claims

1. A speech recognition device comprising:

a communication module receiving speech data corresponding to speech input from a speech recognition terminal and multi-sensor data corresponding to input environment of the speech;

a model selection module selecting a language and acoustic model corresponding to the multi-sensor data among a plurality of language and acoustic models classified according to the speech input environment on the basis of previous multi-sensor data; and

a speech recognition module controlling the communication module to apply a feature vector extracted from the speech data to the language and acoustic model and transmit speech recognition result for the speech data to the speech recognition terminal.

2. The speech recognition device of claim 1, wherein the multi-sensor data comprises at least one of location data for location and position of the speech recognition terminal, image data for a speaker who inputs the speech, and proximity data for the distance between the speech recognition terminal and the speaker.

3. The speech recognition device of claim 2, wherein the model selection module comprises:

a first estimation unit estimating location and position of the speech recognition terminal on the basis of the location data;

a second estimation unit estimating age of the speaker on the basis of the image data;

a third estimation unit estimating signal-to-noise ratio (SNR) for the speech data on the basis of the proximity data; and

a selection unit selecting the language and acoustic model on the basis of the location and position, the age and the signal-to-noise ratio estimated from the first to the third estimation unit.

4. The speech recognition device of claim 2, wherein the model selection module further comprises a database storing the plurality of language and acoustic models classified according to the speech input environment on the basis of the previous multi-sensor data.

5. The speech recognition device of claim 1, wherein the speech recognition module comprises:

a preprocessing unit extracting the feature vector from the speech data; and

a speech recognition unit controlling the communication module to apply the feature vector to the language and acoustic model and transmit speech recognition result for the speech data to the speech recognition terminal.

6. The speech recognition device of claim 5, wherein the preprocessing unit extracts the feature vector by classifying the speech data into frame units and eliminating frame-basis noise components.

7. A method for operating a speech recognition device comprising:

receiving speech data corresponding to speech input from a speech recognition terminal and multi-sensor data corresponding to input environment of the speech;

selecting a language and acoustic model corresponding to the multi-sensor data among a plurality of language and acoustic models classified according to the speech input environment on the basis of previous multi-sensor data; and

transmitting speech recognition result for the speech data to the speech recognition terminal obtained by applying a feature vector extracted from the speech data to the language and acoustic model.

8. The method for operating a speech recognition device of claim 7, wherein the multi-sensor data comprises at least one of location data for location and position of the speech recognition terminal, image data for a speaker who inputs the speech, and proximity data for the distance between the speech recognition terminal and the speaker.

9. The method for operating a speech recognition device of claim 7, wherein the selection step comprises:

estimating location and position of the speech recognition terminal on the basis of the location data;

estimating age of the speaker on the basis of the image data;

estimating signal-to-noise ratio (SNR) for the speech data on the basis of the proximity data; and

selecting the language and acoustic model on the basis of the location and position, the age and the signal-to-noise ratio estimated from the first estimating step to the third estimating step.

10. The method for operating a speech recognition device of claim 7, wherein the transmitting step comprises:

extracting the feature vector from the speech data; and

transmitting speech recognition result for the speech data to the speech recognition terminal obtained by applying the feature vector to the language and acoustic model.

11. The method for operating a speech recognition device of claim 7, wherein the extracting step classifies the speech data into frame units and eliminates frame-basis noise components.