Apparatus, method, and computer program product for speech recognition

A speech recognition apparatus comprising:

• a sound information acquiring unit that acquires sound information;
• a speech recognizing unit that performs a speech recognition of a sound signal;
• a recognition candidate storing unit that stores recognition candidates which are targets of the speech recognition;
• a recognition candidate extracting unit that extracts a predetermined recognition candidate from the recognition candidates stored in the recognition candidate storing unit based on a result of the speech recognition;
• a segment dividing unit that divides the extracted recognition candidate into plural unit segments;
• a segment information acquiring unit that acquires segment information indicating a feature of each unit segment;
• a segment relation value calculating unit that calculates a segment relation value indicating a relative feature of a target segment which is a unit segment to be processed with respect to an adjacent segment which is a unit segment adjacent to the target segment based on segment information of the target segment and segment information of the adjacent segment among the segment information;
• and
• a recognition result selecting unit that selects a recognition result from the recognition candidates based on the segment relation value.

A speech recognition apparatus comprising:

• a sound information acquiring unit that acquires sound information;
• a recognition candidate storing unit that stores plural recognition candidates that are targets of speech recognition;
• a recognition candidate extracting unit that extracts one or more recognition candidates from the plural recognition candidates with the use of a partial sound information which is a portion of the sound information;
• a unit segment dividing unit that divides a portion of the recognition candidate extracted by the recognition candidate extracting unit into plural unit segments;
• a segment information acquiring unit that acquires segment information indicating a feature of each unit segment;
• a segment relation value calculating unit that calculates a segment relation value indicating a relative feature of a target segment which is a unit segment to be processed with respect to an adjacent segment, which is a unit segment adjacent to the target segment and is a segment included in the sound information already acquired, based on segment information of the target segment and segment information of the adjacent segment among the segment information;
• and
• a recognition result selecting unit that selects a recognition result from the recognition candidates with the use of the segment relation value.

A speech recognition apparatus comprising:

• a sound information acquiring unit that acquires sound information;
• a sound segment estimating unit that estimates a sound segment which is a segment corresponding to an utterance of a person from the sound information;
• a duration estimating unit that estimates a duration of the estimated sound segment;
• a recognition candidate storing unit that stores recognition candidates that are targets of speech recognition;
• a recognition candidate extracting unit that extracts one or more recognition candidates from the recognition candidates stored in the recognition candidate storing unit with the use of the estimated sound segment;
• a statistic model storing unit that stores a statistic model which can be taken by a segment relation value indicating a relative feature between two unit segments included in the sound segment;
• a likelihood calculating unit that calculates a likelihood of the recognition candidate extracted by the recognition candidate extracting unit with the use of the statistic model and a duration of the estimated sound segment;
• and
• a recognition result selecting unit that selects a recognition result from the recognition candidates with the use of the likelihood.

A method of speech recognition comprising:

• acquiring sound information;
• dividing the sound information into plural unit segments;
• acquiring segment information that indicates a feature of each unit segment;
• calculating a segment relation value that indicates a relative feature of a target segment which is a unit segment to be processed with respect to an adjacent segment which is a unit segment adjacent to the target segment, based on segment information of the target segment and segment information of the adjacent segment among the segment information;
• and
• selecting a recognition result from recognition candidates stored in a recognition candidate storing unit that stores recognition candidates that are targets of speech recognition, utilizing the segment relation value.

A method of speech recognition comprising:

• acquiring sound information;
• performing a speech recognition of a sound signal;
• extracting a predetermined recognition candidate from recognition candidates stored in a recognition candidate storing unit that stores recognition candidates which are targets of the speech recognition based on a result of the speech recognition;
• dividing the extracted recognition candidate into plural unit segments;
• acquiring segment information indicating a feature of each unit segment;
• calculating a segment relation value indicating a relative feature of a target segment which is a unit segment to be processed with respect to an adjacent segment which is a unit segment adjacent to the target segment based on segment information of the target segment and segment information of the adjacent segment among the segment information;
• and
• selecting a recognition result from the recognition candidates based on the segment relation value.

The method of speech recognition comprising:

• acquiring sound information;
• extracting one or more recognition candidates from plural recognition candidates stored in a recognition candidate storing unit that stores the plural recognition candidates that are targets of speech recognition, with the use of a partial sound information which is a portion of the sound information;
• dividing a portion of the extracted recognition candidate into plural unit segments;
• acquiring segment information indicating a feature of each unit segment;
• calculating a segment relation value indicating a relative feature of a target segment which is a unit segment to be processed with respect to an adjacent segment, which is a unit segment adjacent to the target segment and is a segment included in the sound information already acquired, based on segment information of the target segment and segment information of the adjacent segment among the segment information;
• and
• selecting a recognition result from the recognition candidates with the use of the segment relation value.

A method of speech recognition comprising:

• acquiring sound information;
• estimating a sound segment which is a segment corresponding to an utterance of a person from the sound information;
• estimating a duration of the sound segment;
• extracting one or more recognition candidates from recognition candidates stored in a recognition candidate storing unit that stores the recognition candidates that are targets of speech recognition, with the use of the sound segment;
• calculating a likelihood of the recognition candidate with the use of a statistic model stored in a statistic model storing unit that stores a statistic model which can be taken by a segment relation value indicating a relative feature between two unit segments included in the sound segment and a duration of the sound segment;
• and
• selecting a recognition result from the recognition candidates with the use of the likelihood.

A computer program product having a computer readable recording medium including programmed instructions for speech recognition on sound information, wherein the instructions, when executed by a computer, cause the computer to perform:

• acquiring sound information;
• dividing the acquired sound information into plural unit segments;
• acquiring segment information that indicates a feature of each unit segment;
• calculating a segment relation value that indicates a relative feature of a target segment which is a unit segment to be processed with respect to an adjacent segment which is a unit segment adjacent to the target segment, based on segment information of the target segment and segment information of the adjacent segment among the segment information;
• and
• selecting a recognition result from recognition candidates stored in a recognition candidate storing unit that stores recognition candidates that are targets of speech recognition, utilizing the segment relation value.

A computer program product having a computer readable recording medium including programmed instructions for speech recognition on sound information, wherein the instructions, when executed by a computer, cause the computer to perform:

• acquiring sound information;
• performing a speech recognition of a sound signal;
• extracting a predetermined recognition candidate from recognition candidates stored in a recognition candidate storing unit that stores recognition candidates which are targets of the speech recognition based on a result of the speech recognition;
• dividing the extracted recognition candidate into plural unit segmentst;
• acquiring segment information indicating a feature of each unit segment;
• calculating a segment relation value indicating a relative feature of a target segment which is a unit segment to be processed with respect to an adjacent segment which is a unit segment adjacent to the target segment based on segment information of the target segment and segment information of the adjacent segment among the segment information;
• and
• selecting a recognition result from the recognition candidates based on the segment relation value.

A computer program product having a computer readable recording medium including programmed instructions for speech recognition on sound information, wherein the instructions, when executed by a computer, cause the computer to perform:

• acquiring sound information;
• extracting one or more recognition candidates from plural recognition candidates stored in a recognition candidate storing unit that stores the plural recognition candidates that are targets of speech recognition, with the use of a partial sound information which is a portion of the sound information;
• dividing a portion of the extracted recognition candidate into plural unit segments;
• acquiring segment information indicating a feature of each unit segment;
• calculating a segment relation value indicating a relative feature of a target segment which is a unit segment to be processed with respect to an adjacent segment, which is a unit segment adjacent to the target segment and is a segment included in the sound information already acquired, based on segment information of the target segment and segment information of the adjacent segment among the segment information;
• and
• selecting a recognition result from the recognition candidates with the use of the segment relation value.

A computer program product having a computer readable recording medium including programmed instructions for speech recognition on sound information, wherein the instructions, when executed by a computer, cause the computer to perform:

• acquiring sound information;
• estimating a sound segment which is a segment corresponding to an utterance of a person from the sound information;
• estimating a duration of the sound segment;
• extracting one or more recognition candidates from recognition candidates stored in a recognition candidate storing unit that stores the recognition candidates that are targets of speech recognition, with the use of the sound segment;
• calculating a likelihood of the recognition candidate with the use of a statistic model stored in a statistic model storing unit that stores a statistic model which can be taken by a segment relation value indicating a relative feature between two unit segments included in the sound segment and a duration of the sound segment;
• and
• selecting a recognition result from the recognition candidates with the use of the likelihood.