Voice recognition apparatus and recording medium storing voice recognition program

• US 8,560,317 B2
• Filed: 09/18/2006
• Issued: 10/15/2013
• Est. Priority Date: 05/18/2006
• Status: Expired due to Fees

• Current Assignee: Fujitsu Limited
• Original Assignee: Fujitsu Limited
• Inventors: Abe, Kenji
• Primary Examiner(s): Godbold, Douglas
• Assistant Examiner(s): Villena, Mark
• Application Number:
  US11/522,422
• Publication Number:
  US 20070271097A1
• Time in Patent Office: 2,584 Days
• Field of Search: 704/255, 704/257, 704/249, 704/251, 704/236, 704/225, 704/233, 704/256, 704/240, 707/3
• US Class Current: 704/251
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 2015/228 of application context

Abstract

A vocabulary dictionary storing unit for storing a plurality of words in advance, a vocabulary dictionary managing unit for extracting recognition target words, a matching unit for calculating a degree of matching with the recognition target words based on an accepted voice, a result output unit for outputting, as a recognition result, a word having a best score from a result of calculating the degree of matching, and an extraction criterion information managing unit for changing extraction criterion information according to a result of monitoring by a monitor control unit are provided. The vocabulary dictionary storing unit further includes a scale information storing unit for storing scale information serving as a scale at the time of extracting the recognition target words, and an extraction criterion information storing unit for storing extraction criterion information indicating a criterion of the recognition target words at the time of extracting the recognition target words. With the change in the extraction criterion information, the vocabulary dictionary managing unit increases or decreases the number of the recognition target words. This makes it possible to improve a recognition performance without the need for a user to carry out a troublesome operation.

Claims

1. A voice recognition apparatus comprising:

a voice input unit for converting an accepted voice into digital voice data;

a voice analyzing unit for converting the digital voice data into a feature value;

a vocabulary dictionary storing unit for storing a plurality of words in advance;

a vocabulary dictionary managing unit for extracting recognition target words, which are targeted for recognition, from the plurality of words stored in the vocabulary dictionary storing unit;

a matching unit for calculating a degree of matching with the recognition target words extracted by the vocabulary dictionary managing unit using the feature value converted by the voice analyzing unit; and

a result output unit for outputting, as a recognition result, a word having a best score from a result of calculating the degree of matching by the matching unit;

wherein the voice recognition apparatus comprises

a scale information storing unit for storing in advance scale information serving as a scale for extracting the recognition target words for each of the plurality of words stored in the vocabulary dictionary storing unit,

an extraction criterion information storing unit for storing in advance extraction criterion information for deciding whether the vocabulary dictionary managing unit uses or not the scale information at the time of extracting the recognition target words,

a monitor control unit for monitoring processing states of at least one of the voice input unit, the matching unit and the result output unit, and for outputting an update signal when detecting a given processing state of at least one of the voice input unit, the matching unit and the result output unit,

an extraction criterion information managing unit for changing the extraction criterion information stored in the extraction criterion information storing unit based on the update signal according to the detected processing state of at least one of the voice input unit, the matching unit and the result output unit from the monitor control unit, and

the monitor control unit further monitors an operation accepted from a user and a measurement unit for measuring a time and outputs the update signal based on at least one of the recognition result of the result output unit, the modification, by the user, of the operation or the input voice, an interval between the voice inputs or the operations by the user and a number of calculating the degree of matching by the matching unit,

with the change in the extraction criterion information, the vocabulary dictionary managing unit switches using or not using the scale information for extracting the recognition target words, thereby increases or decreases the number of the recognition target words to be extracted from the plurality of words stored in the vocabulary dictionary storing unit.

2. The voice recognition apparatus according to claim 1, wherein the monitor control unit monitors at least one monitor target from a processing state of an operation accepted from a user, a processing state of the voice accepted from the user and a predetermined time passed from a measurement unit for measuring a time,

the vocabulary dictionary managing unit increases or decreases the number of the recognition target words to be extracted from the plurality of words stored in the vocabulary dictionary storing unit according to the changed extraction criterion information, and

the matching unit calculates the degree of matching with the recognition target words increased or decreased in number using the feature value converted by the voice analyzing unit.

3. The voice recognition apparatus according to claim 2, wherein the monitor control unit monitors at least one monitor target from (a) to (h) below,

when the monitor control unit detects a state of (a) to (d) below, it transmits a first updating signal to the extraction criterion information managing unit, and the extraction criterion information managing unit changes the extraction criterion information stored in the extraction criterion information storing unit based on the first updating signal so that the number of the recognition target words to be extracted from the plurality of words stored in the vocabulary dictionary storing unit increases, and

after transmitting the first updating signal to the extraction criterion information managing unit, when the monitor control unit detects a state of (e) to (h) below, it transmits a second updating signal to the extraction criterion information managing unit, and the extraction criterion information managing unit changes the extraction criterion information stored in the extraction criterion information storing unit based on the second updating signal so that the number of the recognition target words to be extracted from the plurality of words stored in the vocabulary dictionary storing unit decreases:

(a) in a case where the recognition result cannot be obtained,

(b) in a case where the user conducts an operation or makes a speech so as to modify or correct an operation or a speech immediately before it,

(c) in a case where the user does not conduct a confirmation operation even after a predetermined period despite a fact that the confirmation operation of a processing is necessary,

(d) in a case where no operation is conducted or no speech is made for a predetermined period,

(e) in a case where the matching unit calculates the degree of matching predetermined times,

(f) in a case where a predetermined time is passed,

(g) in a case where the user conducts a confirmation operation within a predetermined period when the confirmation operation of the processing is necessary,

(h) in a case where the user does not conduct the operation or does not make the speech so as to modify or correct the operation or the speech immediately before it.

4. The voice recognition apparatus according to claim 2, wherein the vocabulary dictionary managing unit selectively performs one of operations (1) and (2) below according to the extraction criterion information stored in the extraction criterion information storing unit:

(1) the vocabulary dictionary managing unit extracts the recognition target words based on the scale information,

(2) the vocabulary dictionary managing unit extracts all the words stored in the vocabulary dictionary storing unit as the recognition target words regardless of the scale information.

5. The voice recognition apparatus according to claim 1, further comprising

a use frequency managing unit for monitoring the number of uses of each of the plurality of words stored in the vocabulary dictionary storing unit and calculating a use frequency of each of the plurality of words,

a use frequency storing unit for storing, as use frequency data, the use frequency calculated by the use frequency managing unit so as to correspond to each of the plurality of . words stored in the vocabulary dictionary storing unit, and

a scale information managing unit for updating the scale information stored in the scale information storing unit using at least the use frequency data stored in the use frequency storing unit.

6. The voice recognition apparatus according to claim 5, further comprising a statistical data obtaining unit for obtaining statistical data regarding the plurality of words from a statistical data storage device storing the statistical data,

wherein the scale information managing unit updates the scale information stored in the scale information storing unit using the use frequency data stored in the use frequency storing unit and the statistical data obtained by the statistical data obtaining unit.

7. The voice recognition apparatus according to claim 5, further comprising

a vocabulary group use frequency managing unit for dividing the plurality of words stored in the vocabulary dictionary storing unit into a plurality of vocabulary groups and calculating a use frequency of each of the vocabulary groups based on the use frequency of each of the plurality of words belonging to the vocabulary group stored in the vocabulary dictionary storing unit,

a vocabulary group use frequency storing unit for storing, as vocabulary group use frequency data, the use frequency of the vocabulary group calculated by the vocabulary group use frequency managing unit so as to correspond to each of the vocabulary groups, and

a threshold storing unit for storing a threshold indicating a criterion of the vocabulary group use frequency data at the time of extracting the recognition target words,

wherein the vocabulary dictionary managing unit selectively performs one of operations (3) and (4) below referring to the threshold stored in the threshold storing unit and the vocabulary group use frequency data stored in the vocabulary group use frequency storing unit according to the extraction criterion information stored in the extraction criterion information storing unit:

(3) for the vocabulary group whose vocabulary group use frequency data are equal to or larger than the threshold, the vocabulary dictionary managing unit extracts all the words belonging to this vocabulary group as the recognition target words regardless of the scale information,

(4) for the vocabulary group whose vocabulary group use frequency data are smaller than the threshold, the vocabulary dictionary managing unit extracts the recognition target words from the words belonging to this vocabulary group based on the scale information.

8. The voice recognition apparatus according to claim 7, wherein the monitor control unit monitors at least one monitor target from a processing state of an operation accepted from a user, a processing state of the voice accepted from the user and a predetermined time passed from a measurement unit for measuring a time,

the voice recognition apparatus further comprises a threshold managing unit for updating the threshold stored in the threshold storing unit according to the result of the monitoring by the monitor control unit, and

the vocabulary dictionary managing unit selectively performs one of the operations (3) and (4) above according to the updated threshold.

9. The voice recognition apparatus according to claim 8, wherein the monitor control unit monitors at least one monitor target from (a) to (h) below,

when the monitor control unit detects a state of (a) to (d) below, it transmits a first updating signal to the threshold managing unit, and the threshold managing unit updates the threshold stored in the threshold storing unit so as to become smaller based on the first updating signal, and

after transmitting the first updating signal to the threshold managing unit, when the monitor control unit detects a state of (e) to (h) below, it transmits a second updating signal to the threshold managing unit, and the threshold managing unit updates the threshold stored in the threshold storing unit so as to become larger based on the second updating signal:

(a) in a case where the recognition result cannot be obtained,

(b) in a case where the user conducts an operation or makes a speech so as to modify or correct an operation or a speech immediately before it,

(c) in a case where the user does not conduct a confirmation operation even after a predetermined period despite a fact that the confirmation operation of a processing is necessary,

(d) in a case where no operation is conducted or no speech is made for a predetermined period,

(e) in a case where the matching unit calculates the degree of matching predetermined times,

(f) in a case where a predetermined time is passed,

(g) in a case where the user conducts a confirmation operation within a predetermined period when the confirmation operation of the processing is necessary,

(h) in a case where the user does not conduct an operation or does not make a speech so as to modify or correct the operation or the speech immediately before it.

10. The voice recognition apparatus according to claim 1, further comprising a voice storing unit for storing the digital voice data converted by the voice input unit or the feature value converted by the voice analyzing unit,

wherein the matching unit calculates the degree of matching with the recognition target words extracted by the vocabulary dictionary managing unit using the digital voice data or the feature value stored in the voice storing unit.

11. The voice recognition apparatus according to claim 10, wherein when the matching unit calculates the degree of matching with the recognition target words extracted by the vocabulary dictionary managing unit using the digital voice data or the feature value stored in the voice storing unit, the vocabulary dictionary managing unit excludes the words extracted as the recognition target words immediately before that and extracts the recognition target words from the plurality of words stored in the vocabulary dictionary storing unit.

12. A non-transitory recording medium storing a voice recognition program causing a computer to execute

a voice input operation of converting an accepted voice into digital voice data;

a voice analyzing operation of converting the digital voice data into a feature value;

a vocabulary dictionary managing operation of extracting recognition target words, which are targeted for recognition, from a vocabulary dictionary storing unit storing a plurality of words in advance;

a matching operation of calculating a degree of matching with the recognition target words extracted in the vocabulary dictionary managing operation using the feature value converted in the voice analyzing operation; and

a result output operation of outputting, as a recognition result, a word having a best score from a result of calculating the degree of matching in the matching operation;

wherein the vocabulary dictionary managing operation causes a computer to execute

an operation of referring to scale information serving as a scale for extracting the recognition target words stored in advance in a scale information storing unit and referring to extraction criterion information for deciding whether the scale information is used or is not used at the time of extracting the recognition target words stored in advance in an extraction criterion information storing unit, and

causes a computer to execute

a monitor control operation of monitoring processing states of at least one of the voice input operation, the matching operation and the result output operation, and outputting an update signal when detecting a given processing state of at least one of the voice input operation, the matching operation and the result output operation,

an extraction criterion information managing operation of changing the extraction criterion information stored in the extraction criterion information storing unit based on the update signal according to the detected processing state of at least one of the voice input operation, the matching operation and the result output operation in the monitor control operation, and

the monitor control operation further includes monitoring an user operation accepted from a user and a measurement operation by the computer for measuring a time and outputting the update signal based on at least one of the recognition result of the result output operation, the modification by the user, of the user operation or the input voice, an interval between the voice inputs or the user operations by the user and a number of calculating the degree of matching by the matching operation,

with the change in the extraction criterion information, the vocabulary dictionary managing operation causes a computer to switch between the operation of the extracting the recognition target words by using the scale information and the operation of the extracting the recognition target word without using the scale information, thereby execute an operation of increasing or decreasing the number of the recognition target words to be extracted from the plurality of words stored in the vocabulary dictionary storing unit.