Speech recognition with plural confidence measures

• US 7,043,429 B2
• Filed: 03/28/2002
• Issued: 05/09/2006
• Est. Priority Date: 08/24/2001
• Status: Expired due to Fees

• Current Assignee: Industrial Technology Research Institute
• Original Assignee: Industrial Technology Research Institute
• Inventors: Chang, Sen-Chia, Tu, Jia-Jang, Chien, Shih-Chien
• Primary Examiner(s): Lerner, Martin
• Application Number:
  US10/107,314
• Publication Number:
  US 20030040907A1
• Time in Patent Office: 1,503 Days
• Field of Search: 704/231, 704/233, 704/236, 704/239, 704/240, 704/251, 704/252, 704/255
• US Class Current: 704/236
• CPC Class Codes: G10L 15/08 Speech classification or search
• CPC Class Codes: G10L 15/32 Multiple recognisers used in sequence or in parallel; Score combination systems therefor, e.g. voting systems

Abstract

A speech recognition system is used to receive a speech signal and output an output language word with respect to the speech signal. The speech recognition system has preset quantities for a first threshold, a second threshold, and a third threshold. The speech recognition system includes a first speech recognition device that is used to receive the speech signal and generate a first candidate language word and a first confidence measurement of the first candidate language word, according to the speech signal. A second speech recognition device is used to receive the speech signal and generate a second candidate language word and a second confidence measurement of the second candidate language word, according to the speech signal. A confidence measurement judging unit is used to output the language word, by comparing the first confidence measurement and the second confidence measurement to the above thresholds.

Claims

1. A speech recognition system, used for receiving a speech signal and output an output language word with respect to the speech signal, wherein the speech recognition system has a first threshold, a second threshold, and a third threshold, the speech recognition system comprising:

a first speech recognition device, used to receive the speech signal and generate a first candidate language word and a first confidence measurement of the first candidate language word, according to the speech signal;

a second speech recognition device, used to receive the speech signal and generate a second candidate language word and a second confidence measurement of the second candidate language word, according to the speech signal; and

a confidence measurement judging unit, used to judge and output the output language word, according to the first confidence measurement and the second confidence measurement;

wherein when the first confidence measurement is less than the first threshold and the second confidence measurement is less than the second threshold, the first candidate language word is taken as the output language word, when the first confidence measurement is greater than the first threshold and the second confidence measurement is less than the third threshold, the first candidate language word is set to be the output language word, when the first confidence measurement is less than the first threshold and the second confidence measurement is greater than the second threshold, then the second candidate language word is set to be the output language word, and when the second confidence measurement is greater than the third threshold, the second candidate language word is set to be the output language word.

2. The speech recognition system according to claim 1, wherein the first speech recognition device is a continuous speech recognition device.

3. The speech recognition system according to claims 1 or 2, wherein the second speech recognition device is an isolated word speech recognition device.

4. The speech recognition system according to claim 3, wherein the second speech recognition device can recognize at least one language.

5. A speech recognition system, used to receive a speech signal and output an output language word with respect to the speech signal, wherein the speech recognition system has preset quantities for a first threshold and a second threshold, the speech recognition system further includes a storage device, wherein the storage device is used to receive the speech signal and output the speech signal, the speech recognition system comprising:

a first speech recognition device, which is used to receive the speech signal and generate a first candidate language word and a first confidence measurement of the first candidate language word, according to the speech signal;

a confidence measurement judging unit, which is used to determine the output language word; and

a second speech recognition device, which is controlled by the confidence measurement judging unit and is used to receive an output of the speech signal output from the storage device and generate a second candidate language word and a second confidence measurement of the second candidate language word, according to the speech signal,

wherein the confidence measurement judging unit judges whether or not the first confidence measurement is greater than the first threshold, if it being yes, then the first candidate language word is taken as the output language word, if it being no, then the confidence measurement judging unit causes the second speech recognition device to generate a second language word and a second confidence measurement, and then judges whether or not the second confidence measurement is greater than the second threshold, if it being yes, then the second candidate language word is taken as the output language word, if it being no, then the first candidate language word is taken as the output language word.

6. The speech recognition system according to claim 5, wherein the first speech recognition device is a continuous speech recognition device.

7. The speech recognition system according to claims 5 or 6, wherein the second speech recognition device is an isolated word speech recognition device.

8. The speech recognition system according to claim 7, wherein the second speech recognition device can recognize at least one language.

9. A speech recognition method, the method comprising the following steps:

feeding a speech signal into a first speech recognition device and a second speech recognition device;

the first speech recognition device generating a first candidate language word and a first confidence measurement, according to the speech signal, and the second speech recognition device generating a second candidate language word and a second confidence measurement, according to the speech signal; and

if the first confidence measurement being less than the first threshold and the second confidence measurement being less than the second threshold, then the first candidate language word being taken as the output language word, if the first confidence measurement being greater than the first threshold and the second confidence measurement being less than a third threshold, then the first candidate language word being taken as the output language word, if the first confidence measurement being less than the first threshold and the second confidence measurement being greater than the second threshold, then the second candidate language word being taken as the output language word, and if the second confidence measurement being greater than the third threshold, then the second candidate language word being taken as the output language word.

10. The speech recognition method according to claim 9, wherein first speech recognition device is a continuous speech recognition device.

11. The speech recognition method according to claim 9 or 10, wherein the second speech recognition device is an isolated word speech recognition device.

12. The speech recognition method according to claim 11, wherein the second speech recognition device can recognize at least one language.

13. A speech recognition method, the method comprising the following steps:

(a) feeding a speech signal into a first speech recognition device;

(b) the first speech recognition device generating a first candidate language word and a first confidence measurement, according to the speech signal;

(c) judging whether or not the first confidence measurement is greater than the first threshold, if it being yes, then the first candidate language word being taken as the output language word and then the method goes to an end;

(d) feeding the speech signal into a second speech recognition device and the second speech recognition device generating a second candidate language word and a second confidence measurement, according to the input speech signal; and

(e) judging whether or not the second confidence measurement is greater than the second threshold, if it being yes, then the second candidate language word being taken as the output language word, if it being no, then the first candidate language word being taken as the output language word.

14. The speech recognition method according to claim 13, wherein first speech recognition device is a continuous speech recognition device.

15. The speech recognition method according to claims 13 or 14, wherein the second speech recognition device is an isolated word speech recognition device.

16. The speech recognition method according to claim 15, wherein the second speech recognition device can recognize at least one language.