Speech recognition system

Claim 2

A speech recognition system comprising: parameter extracting means (10) for analyzing input speech and extracting a speech parameter of the input speech; first word recognizing means (20), based on matching in units of words, for segmenting the speech parameter extracted by said parameter extracting means (10) in units of words so as to extract a word speech pattern and for performing word recognition by matching the word speech pattern with a word reference pattern; second word recognizing means (30), based on matching in units of word constituent elements, for performing recognition of a word constituent element by matching a speech pattern of each. word constituent element using the speech parameter extracted by said parameter extracting means (10) with a prestored word constituent element reference pattern, and for performing word recognition on the basis of the obtained word constituent element series candidates; recognition result output means (40) for totally obtaining a recognition result on the basis of word recognition results from said first and second word recognizing means (20, 30) and outputting the obtained recognition result; and learning means (50) for performing learning of the word reference pattern on the basis of the recognition result obtained by said recognition result output means (40) and a word speech pattern extracted in the course of the recognition processing.

Claim 3

A system according to claim 1, characterized in that said first word recognizing means (20) includes storage means (23) for storing a word reference pattern, and said learning means (50) includes means (52) for updating a storage content of said storage means.

Claim 4

A system according to claim 1, characterized in that said second word recognizing means (30) uses phonemes as units of a word constituent element.

Claim 5

A system according to claim 1, characterized in that said second word recognizing means (30) uses syllables as units of a word constituent element.

Claim 6

A system according to claim 1, characterized in that said recognition result output means (40) includes means (41) for obtaining a total recognition result by weighting the recognition results obtained by said first and second word recognizing means (20, 30) in accordance with characteristics of said first and second word recognizing means (20,30), respectively.

Claim 7

A system according to claim 1, characterized in that said learning means (50) includes input means (43) for allowing a user to designate a category name of the recognized input speech as a learning word speech pattern and means (52) for collecting the word speech patterns in response to a designation input through said input means (43).

Claim 8

A system according to claim 1, characterized in that said learning means (50) includes means (50) for storing non-learned recognition information in said first and second word recognizing means (20, 30) and said recognition result output means (40), counting means (52) for counting the number of word speech patterns of the non-learned recognition information, and means (52) for executing learning on the basis of the recognition information in response to said counting means every time the number of data reaches a predetermined value.

Claim 9

A speech recognition system comprising: parameter extracting means (10) for analyzing input speech and extracting a speech parameter of the inputspeech; first word recognizing means (20), based on matching in units of words, for segmenting the speech parameter extracted by said parameter extracting means (10) in units of words so as to extract a word speech pattern and for performing word recognition by matching the word speech pattern with a word reference pattern; second word recognizing means (30), based on matching in units of word constituent elements, for performing recognition of a word constituent element by matching a speech pattern of each word constituent element using the speech parameter extracted by said parameter extracting means (10) with a prestored word constituent element reference pattern, and for performing word recognition on the basis of the obtained word constituent element series candidates; recognition result output means (40) for totally obtaining a recognition result on the basis of word recognition results from said first and second word recognizing means (20, 30) and outputting the obtained recognition result; and learning means (50) for performing learning of the word reference pattern on the basis of the recognition result obtained by said recognition result output means (40) and a word speech pattern extracted in the course of the recognition processing, wherein said recognition output means (40) changes a ratio of contributions of a word recognition result obtained by said first word recognizing means (20) based on matching in units of words and of a word recognition result obtained by said second word recognizing means (30) based on matching in units of word constituent elements in the finally output total recognition result in accordance with the progress of learning of said learning means (50), and increases a priority of the word recognition result obtained by said first word recognizing means (20) with respect to that of the word recognition result obtained by said second word recognizing means (30) with the progress of the learning.

Claim 10

A speech recognition system comprising: parameter extracting means (10) for analyzing input speech and extracting a speech parameter of the inputspeech; first word recognizing means (20), based on matching in units of words, for segmenting the speech parameter extracted by said parameter extracting means (10) in units of words so as to extract a word speech pattern and for performing word recognition by matching the word speech pattern with a word reference pattern; second word recognizing means (30), based on matching in units of word constituent elements, for performing recognition of a word constituent element by matching a speech pattern of each word constituent element using the speech parameter extracted by said parameter extracting means (10) with a prestored word constituent element reference pattern, and for performing word recognition on the basis of the obtained word constituent element series candidates; recognition result output means (40) for totally obtaining a recognition result on the basis of word recognition results from said first and second word recognizing means (20, 30) and outputting the obtained recognition result; and learning means (50) for performing learning of the word reference pattern on the basis of the recognition result obtained by said recognition result output means (40) and a word speech pattern extracted in the course of the recognition processing, wherein said recognition output means (40) changes a ratio of contributions of a word recognition result obtained by said first word recognizing means (20) based on matching in units of words and of a word recognition result obtained by said second word recognizing means (30) based on matching in units of word constituent elements in the finally output total recognition result in accordance with the progress of learning of said learning means (50), and increases a priority of the word recognition result obtained by said first word recognizing means (20) with respect to that of the word recognition result obtained by said second word recognizing means (30) with the progress of the learning.

Claim 11

A system according to claim 8, characterized in that said recognition result output means (40) includes progress discriminating means (41) for discriminating progress of learning of said learning means (50), and contribution control means (41) for controlling contributions of recognition results, which are obtained by said first and second word recognizing means (20, 30) to obtain a final recognition result, so as to shift a priority from a discrimination result obtained by said second word recognizing means (30) to a discrimination result obtained by said first word recognizing means (20) with the progress of learning.

Claim 12

A system according to claim 9, characterized in that said progress discriminating means (41) includes accumulation value discriminating means (41) for discriminating an accumulation value of the number of word speech pattern supplied for learning of said learning means (50).

Claim 13

A system according to claim 9, characterized in that said contribution control means (41) includes means (41) for controlling priorities of recognition results which are obtained by said first and second word recognizing means (20, 30) so as to obtain a final recognition result.

Claim 14

A system according to claim 9, characterized in that said contribution control means (41) includes means (41) for selecting a processing algorithm from a plurality of processing algorithms having substantially different priorities for recognition results, which are obtained by said first and second word recognizing means (20, 30) so as to obtain a final recognition result, in accordance with a discrimination result obtained by said progress discriminating means (41).

Claim 15

A system according to claim 8, characterized in that said first word recognizing means (20) includes storage means (23) for storing a word reference pattern and said learning means (50) includes means (52) for updating a storage content of said storage means.

Claim 16

A system according to claim 8, characterized in that said second word recognizing means (30) uses phonemes as units of a word constituent element.

Claim 17

A system according to claim 8, characterized in that said second word recognizing means (30) uses syllables as units of a word constituent element.

Claim 18

A system according to claim 8, characterized in that said recognition result output means (40) includes means (41) for obtaining a total recognition result by weighting the recognition results obtained by said first and second word recognizing means (20, 30) in accordance with characteristics of said first and second wrod recognizing means (20,30), respectively.

Claim 19

A system according to claim 8, characterized in that said learning means (50) includes input means (43) for allowing a user to designate a category name of the recognized input speech as a learning word speech pattern and means (52) for collecting the word speech patterns in response to a designation input through said input means (43).

Claim 20

A system according to claim 8, characterized in that said learning means (50) includes means (50) for storing recognition information in said first and second word recognizing means (20, 30) and said recognition result output means (40), counting means (41) for counting an accumulation value of the number of word speech patterns subjected to recognition processing, and means (52) for executing learning on the basis of the recognition information in response to said counting means (41) every time the accumulation value reaches a predetermined value.