System and method for performing compensated speech recognition

A speech recognition system comprising:

• A processor executing an application for:
• prompting a user to speak a first utterance, the first utterance comprising at least one phrase;
• in response to a determination that a first confidence value representing an expected accuracy of a speech recognition operation on the first utterance is less than a threshold value, prompting the user to speak a second utterance, wherein the second utterance comprises the at least one phrase associated with the first utterance;
• and
• recording the first and second utterances;
• a difference engine for processing the first and second utterances to identify a plurality of acoustic differences between first and second utterances;
• a speech recognition engine for:
• performing speech recognition on the first utterance and generating a first recognition result, wherein the first recognition result comprises the first confidence value;
• and
• performing speech recognition on the second utterance and generating a second recognition result, wherein the second recognition result comprises a second confidence value
• a compensation controller for compensating the speech recognition system by adjusting a plurality of adjustable parameters to compensate for the plurality of acoustic differences;
• wherein the speech recognition engine further processes the first and/or the second user utterances by performing compensated speech recognition of the first and/or the second user utterances, wherein the performing compensated speech recognition comprises:
• performing speech recognition on a selected one of the first user utterance and the second user utterance and generating a compensated speech recognition result for the selected utterance using the compensated speech recognition system, wherein the first user utterance is selected when the first confidence value is higher than the second confidence value, and wherein the second user utterance is selected when the second confidence value is higher than the first confidence value;
• or
• performing speech recognition on the first user utterance to generate a first compensated result having a first compensated confidence value using the compensated speech recognition system, performing speech recognition on the second user utterance to generate a second compensated result having a second compensated confidence value using the compensated speech recognition system, selecting the first compensated result as the compensated speech recognition result when the first compensated confidence value is greater than the second compensated confidence value, and selecting the second compensated result as the compensated speech recognition result when the second compensated confidence value is greater than the first compensated confidence value.

A computer program product comprising a non-transitory computer readable recording medium having computer readable code stored thereon, said computer readable code when executed on a computer system causes the computer system to perform a voice recognition method comprising acts of:

• prompting a user to provide a first utterance comprising a phrase;
• recording a first user utterance;
• performing speech recognition on the first user utterance and generating a first recognition result using a speech recognition system, wherein the first recognition result comprises a first confidence value that represents an expected accuracy of the speech recognition on the first user utterance;
• in response to a determination that the first confidence value is less than a threshold value, re-prompting the user to provide a second utterance comprising the same phrase as the first utterance;
• recording the second user utterance uttered in response to the re-prompting;
• performing speech recognition on the second user utterance and generating a second recognition result using the speech recognition system, wherein the second recognition result comprises a second confidence value that represents an expected accuracy of the speech recognition on the second user utterance;
• in response to a determination that the second confidence value is less than a threshold value, detecting a plurality of acoustic differences between the first and second user utterances;
• modifying, by compensating, the speech recognition system by adjusting a plurality of adjustable parameters to compensate for the plurality of acoustic differences;
• and
• performing compensated speech recognition on the first and/or the second user utterances, wherein the performing compensated speech recognition comprises:
• performing speech recognition on a selected one of the first user utterance and the second user utterance and generating a compensated speech recognition result for the selected utterance using the compensated speech recognition system, wherein the first user utterance is selected when the first confidence value is higher than the second confidence value, and wherein the second user utterance is selected when the second confidence value is higher than the first confidence value;
• or
• performing speech recognition on the first user utterance to generate a first compensated result having a first compensated confidence value using the compensated speech recognition system, performing speech recognition on the second user utterance to generate a second compensated result having a second compensated confidence value using the compensated speech recognition system, selecting the first compensated result as the compensated speech recognition result when the first compensated confidence value is greater than the second compensated confidence value, and selecting the second compensated result as the compensated speech recognition result when the second compensated confidence value is greater than the first compensated confidence value.