Adapting speech recognition acoustic models with environmental and social cues

• US 9,208,781 B2
• Filed: 04/05/2013
• Issued: 12/08/2015
• Est. Priority Date: 04/05/2013
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Bell, Denise A, DeLuca, Lisa Seacat, Jenkins, Jana H, Kusnitz, Jeffrey A
• Primary Examiner(s): He, Jialong
• Assistant Examiner(s): WANG, YI SHENG
• Application Number:
  US13/857,571
• Publication Number:
  US 20140303970A1
• Time in Patent Office: 977 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G10L 15/00 Speech recognition G10L17/00 takes precedence
• CPC Class Codes: G10L 15/20 Speech recognition techniques specially adapted for robustness in adverse environments, e.g. in noise, of stress induced speech G10L21/02 takes precedence
• CPC Class Codes: G10L 15/24 Speech recognition using non-acoustical features

Abstract

An acoustic model adaptation system includes a memory device and a model selector engine coupled to the memory device. The model selector engine is configured to compile information of environmental conditions to identify a current speech environment for audio input into a speech recognizer on a device. The model selector engine is further configured to compare the information of the environmental conditions with profiles of acoustic models. Each profile associates with an acoustic model. Each acoustic model compensates for background noise or acoustical distortions of the audio input. The model selector engine is further configured to select a first acoustic model for the speech recognizer based on the information of the environmental conditions exclusive of audio input from the user.

Claims

1. A computer program product comprising:

a non-transitory computer useable storage medium including a computer readable program, wherein the computer readable program includes instructions that, when executed by a processing device within a computer, cause the computer to perform operations comprising;

compiling information of environmental conditions, exclusive of audio input, to identify a current speech environment for audio input into a speech recognizer on a device, wherein compiling the information of the environmental conditions comprises determining an estimate of a number of persons within a certain distance of the device, wherein determining the estimate of the number of persons within the certain distance of the device comprises compiling information of a number of local wireless devices within the distance of the device;

comparing the information of the environmental conditions with profiles of acoustic models, wherein each profile associates with an acoustic model, and wherein each acoustic model compensates for background noise or acoustical distortions of the audio input; and

selecting a first acoustic model for the speech recognizer based on the information of the environmental conditions exclusive of audio input.

2. The computer program product of claim 1, wherein the operations further comprise creating a new acoustic model and a new profile in response to detecting acoustic characteristics of the audio input different from the selected first profile by a predetermined threshold.

3. The computer program product of claim 2, wherein the operations further comprise adjusting the new acoustic model to match the acoustic characteristics of the audio input.

4. The computer program product of claim 1, wherein determining the estimate further comprises compiling information of social network statistics.

5. The computer program product of claim 1, wherein compiling the information of the environmental conditions comprises identifying a location of the device and comparing the location of the device with a location database.

6. The computer program product of claim 1, wherein compiling the information of the environmental conditions comprises detecting a microphone connected to the device to determine a location of the device.

7. A method for adapting speech recognition acoustic models, the method comprising:

compiling information of environmental conditions, exclusive of audio input, to identify a current speech environment for audio input into a speech recognizer on a device, wherein compiling the information of the environmental conditions comprises determining an estimate of a number of persons within a certain distance of the device, wherein determining the estimate comprises compiling information of social network statistics;

comparing the information of the environmental conditions with profiles of acoustic models, wherein each profile associates with an acoustic model, and wherein each acoustic model compensates for background noise or acoustical distortions of the audio input; and

selecting a first acoustic model for the speech recognizer based on the information of the environmental conditions exclusive of audio input.

8. The method of claim 7, wherein the method further comprises creating a new acoustic model and a new profile in response to detecting acoustic characteristics of the audio input different from the selected first profile by a predetermined threshold.

9. The method of claim 8, wherein the method further comprises adjusting the new acoustic model to match the acoustic characteristic of the audio input.

10. The method of claim 7, wherein determining the estimate further comprises compiling information of a number of local wireless devices within the distance of the device.

11. The method of claim 7, wherein compiling the information of the environmental conditions comprises identifying a location of the device and comparing the location of the device with a location database.

12. The method of claim 7, wherein compiling the information of the environmental conditions comprises detecting a microphone connected to the device to determine a location of the device.

13. An acoustic model adaptation system, the system comprising:

a memory device, and

a model selector engine coupled to the memory device, the model selector engine configured to;

compile information of environmental conditions, exclusive of audio input from a user, to identify a current speech environment for audio input into a speech recognizer on a device;

compare the information of the environmental conditions with profiles of acoustic models, wherein each profile associates with an acoustic model, and wherein each acoustic model compensates for background noise or acoustical distortions of the audio input;

select a first acoustic model for the speech recognizer based on the information of the environmental conditions exclusive of audio input from the user; and

determine an estimate of a number of persons within a certain distance of the device by compilation of information of a number of local wireless devices.

14. The system of claim 13, wherein the system further comprises an adaptor engine coupled to the memory device, the adaptor engine configured to create a new acoustic model and a new profile in response to detection of acoustic characteristics of the audio input different from the selected first profile by a predetermined threshold.

15. The system of claim 14, wherein the adaptor engine is further configured to adjust the new acoustic model to match the acoustic characteristics of the audio input.

16. The system of claim 13, wherein the model selector engine is further configured to determine an estimate of a number of persons within a certain distance of the device by compilation of information of social network statistics.

17. The system of claim 13, wherein the model selector engine is further configured to identify a location of the device and compare the location of the device with a location database.