DYNAMIC SPEECH SYSTEM TUNING

• US 20150248881A1
• Filed: 03/03/2014
• Published: 09/03/2015
• Est. Priority Date: 03/03/2014
• Status: Active Grant

• Current Assignee: General Motors LLC (General Motors Company)
• Original Assignee: General Motors LLC (General Motors Company)
• Inventors: Holdren, John L., Talwar, Gaurav, Zhao, Xufang
• Granted Patent: US 9,911,408 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G10L 13/00 Speech synthesis; Text to speech systems
• CPC Class Codes: G10L 13/08 Text analysis or generation of parameters for speech synthesis out of text, e.g. grapheme to phoneme translation, prosody generation or stress or intonation determination
• CPC Class Codes: G10L 15/01 Assessment or evaluation of speech recognition systems
• CPC Class Codes: G10L 15/06 Creation of reference templates; Training of speech recognition systems, e.g. adaptation to the characteristics of the speaker's voice G10L15/14 takes precedence

Abstract

A system and method of tuning speech recognition systems includes performing text-to-speech conversion of text data; detecting the accuracy of speech converted from text data; determining that the detected accuracy is below a predetermined threshold; recording a user recitation of the text data in response to the determination; and storing the user recitation in an exception database located at a vehicle.

Claims

1. A method of tuning speech recognition systems, comprising the steps of:

(a) performing text-to-speech conversion of text data;

(b) detecting the accuracy of speech converted from text data;

(c) determining that the detected accuracy is below a predetermined threshold;

(d) recording a user recitation of the text data in response to the determination in step (c); and

(e) storing the user recitation in an exception database located at a vehicle.

2. The method of claim 1, further comprising the step of accessing the exception database using an automatic speech recognition (ASR) system.

3. The method of claim 1, further comprising the step of identifying the text data in response to a selection made on a visual display.

4. The method of claim 1, further comprising the step of incrementing a counter each time the text data is converted to speech.

5. The method of claim 1, wherein the accuracy is detected based at least partly on input received from a vehicle occupant.

6. The method of claim 1, wherein the predetermined threshold lies above a medium confidence result and a low confidence result.

7. The method of claim 1, further comprising the steps of linking the user recitation with the text data.

8. A method of tuning speech recognition systems, comprising the steps of:

(a) performing text-to-speech conversion of text data at a vehicle;

(b) presenting the text data as converted speech via an audio system at the vehicle;

(c) detecting the accuracy of the speech converted from text data;

(d) determining that the detected accuracy is below a predetermined threshold;

(e) prompting a vehicle occupant for a recitation of the text data;

(f) recording the recitation of the text data as an audio file; and

(g) adding the audio file to an exception database maintained at the vehicle.

9. The method of claim 8, further comprising the step of accessing the exception database using an automatic speech recognition (ASR) system.

10. The method of claim 8, further comprising the step of identifying the text data in response to a selection made on a visual display.

11. The method of claim 8, further comprising the step of incrementing a counter each time the text data is converted to speech.

12. The method of claim 8, wherein the accuracy is detected based at least partly on input received from the vehicle occupant.

13. The method of claim 8, wherein the predetermined threshold lies above a medium confidence result and a low confidence result.

14. The method of claim 8, further comprising the steps of linking the audio file with the text data.

15. A method of tuning speech recognition systems, comprising the steps of:

(a) receiving speech input at a microphone from a user;

(b) recording the received speech input received from the user;

(c) determining using an automatic speech recognition (ASR) system that the recognition of speech input fell below a high-confidence threshold;

(d) separating the recorded speech into sounds in response to step (c); and

(e) associating the separated sounds with the received speech in an exception dictionary that is accessed by both the ASR system and a text-to-speech (TTS) system.

16. The method of claim 15, wherein the recorded speech input is received and stored at a vehicle.

17. The method of claim 15, further comprising the step of determining that the recognition of speech input fell below a confidence threshold more than a predetermined number of times.

18. The method of claim 15, wherein the sounds are associated with a word represented by text data.