System and method for handling missing speech data
View Patent ↗Disclosed herein are systems, computer-implemented methods, and tangible computer-readable media for handling missing speech data. The computer-implemented method includes receiving speech with a missing segment, generating a plurality of hypotheses for the missing segment, identifying a best hypothesis for the missing segment, and recognizing the received speech by inserting the identified best hypothesis for the missing segment. In another method embodiment, the final step is replaced with synthesizing the received speech by inserting the identified best hypothesis for the missing segment. In one aspect, the method further includes identifying a duration for the missing segment and generating the plurality of hypotheses of the identified duration for the missing segment. The step of identifying the best hypothesis for the missing segment can be based on speech context, a pronouncing lexicon, and/or a language model. Each hypothesis can have an identical acoustic score.
1. A method comprising:
receiving, at a computing device, speech represented by speech packets transmitted over a packet network;
generating, via a processor of the computing device, a plurality of hypotheses for a missing segment of the speech, the plurality of hypotheses based on an acoustic feature of the speech, wherein the missing segment is missing due to lost packets of the speech as the speech packets are transmitted over the packet network;
assigning a same acoustic score to each hypothesis in the plurality of hypotheses;
evaluating the plurality of hypotheses using the acoustic score to identify a possible hypothesis that represents the missing segment of the speech; and
outputting a modified version of the speech, wherein the modified version of the speech is constructed by inserting the possible hypothesis into the missing segment of the speech.
2. The method of claim 1 , wherein the evaluating of the plurality of hypotheses is further based on a context of the speech.
3. The method of claim 2 , wherein generating of the plurality of hypotheses further comprises:
identifying a duration for the missing segment; and
generating the plurality of hypotheses having a length equal to the duration.
4. The method of claim 1 , wherein evaluating of the plurality of hypotheses is based on a pronouncing lexicon.
5. The method of claim 1 , wherein evaluating of the plurality of hypotheses is based on a language model.
6. The method of claim 1 , wherein the same acoustic score represents a certainty of the plurality of hypotheses.
7. The method of claim 1 , wherein evaluating of the plurality of hypotheses is based on a duration model.
8. The method of claim 7 , wherein the duration model is based on a mean and a variance of duration for individual context-dependent phoneme acoustic models associated with the plurality of hypotheses.
9. A system comprising:
a processor; and
a computer-readable storage medium having instructions stored which, when executed by the processor, cause the processor to perform operations comprising:
receiving speech represented by speech packets transmitted over a packet network;
generating a plurality of hypotheses for a missing segment of the speech, the plurality of hypotheses based on an acoustic feature of the speech, wherein the missing segment is missing due to lost packets of the speech as the speech packets are transmitted over the packet network;
assigning a same acoustic score to each hypothesis in the plurality of hypotheses;
evaluating the plurality of hypotheses using the acoustic score to identify a possible hypothesis that represents the missing segment of the speech; and
outputting a modified version of the speech, wherein the modified version of the speech is constructed by inserting the possible hypothesis into the missing segment of the speech.
10. The system of claim 9 , wherein the evaluating of the plurality of hypotheses is further based on a context of the speech.
11. The system of claim 10 , wherein generating of the plurality of hypotheses further comprises:
identifying a duration for the missing segment; and
generating the plurality of hypotheses having a length equal to the duration.
12. The system of claim 9 , wherein evaluating of the plurality of hypotheses is based on a pronouncing lexicon.
13. The system of claim 9 , wherein evaluating of the plurality of hypotheses is based on a language model.
14. The system of claim 9 , wherein the same acoustic score represents a certainty of the plurality of hypotheses.
15. The system of claim 9 , wherein evaluating of the plurality of hypotheses is based on a duration model.
16. The system of claim 15 , wherein the duration model is based on a mean and a variance of duration for individual context-dependent phoneme acoustic models associated with the plurality of hypotheses.
17. A computer-readable storage device having instructions stored which, when executed by a computing device, cause the computing device to perform operations comprising:
receiving, at the computing device, speech represented by speech packets transmitted over a packet network;
generating, via a processor of the computing device, a plurality of hypotheses for a missing segment of the speech, the plurality of hypotheses based on an acoustic feature of the speech, wherein the missing segment is missing due to lost packets of the speech as the speech packets are transmitted over the packet network;
assigning a same acoustic score to each hypothesis in the plurality of hypotheses;
evaluating the plurality of hypotheses using the acoustic score to identify a possible hypothesis that represents the missing segment of the speech; and
outputting a modified version of the speech, wherein the modified version of the speech is constructed by inserting the possible hypothesis into the missing segment of the speech.
18. The computer-readable storage device of claim 17 , wherein the evaluating of the plurality of hypotheses is further based on a context of the speech.
19. The computer-readable storage device of claim 18 , wherein generating of the plurality of hypotheses further comprises:
identifying a duration for the missing segment; and
generating the plurality of hypotheses having a length equal to the duration.
20. The computer-readable storage device of claim 19 , wherein evaluating of the plurality of hypotheses is based on a pronouncing lexicon.