EP0200347A1
Expired / Lapsed
EP
⏱️ 0.0 years remaining
Knowledge-guided automatic speech recognition apparatus and method
Assignee: Toshiba Corp
•
Filing: 1986-03-25
•
Granted: —
•
Expires: 2006-03-25
Abstract
Abstract An acoustic pattern of continuous input speech is divided by an acoustic analyzer (10) into frames of a predetermined time interval. A similarity calculator (12) calculates similarities between frame data and the reference phonemic labels prestored in a dictionary memory (14), and supplies similarity data to a main processor (16), which has memories (22 and 24) prestores speech duration data and connectability data. The main processor (16) extracts, from among the reference phonemic labels, those which satisfy phonetic/phonological conditions with respect to phonemes of the input speech. Similarity sum calculation is conducted only for the similarity data of the extracted labels.
Inventors
- 👤 Teruhiko C/O Patent Division Ukita
Bibliographic Details
Application Number
—
Jurisdiction
EP
Data Source
google_patents
Last Synchronized
2026-10-28 12:39
CPC Classifications
G; G10; G10L; G10L15/00
Claims Analysis
Total Claims: 14
Claim 1
Independent Claim
182 words
An acoustic pattern recognition apparatus for automatically recognizing continuous input speech, comprising acoustic analysis means (10) for dividing an acoustic pattern of the continous input speech at predetermined time intervals so as to produce a plurality of frame data, dictionary memory means (14) for storing reference acoustic patterns of phonemes in a selected language as reference phonemic labels, and similarity calculation means (12) for calculating similarities between the frame data and the reference acoustic patterns so as to produce a plurality of similarity data, characterized in that main processor means (16) is connected to said similarity calculation means, said main processor means prestoring at least one phonetic/ phonological data as a condition in the processing of phonemic labels, extracting, from among the reference phonemic labels to be compared with the frame data, reference phonemic labels which satisfy the condition for the phonetic/phonological data with respect to the phonemes of the input speech, and calculating similarity sums for the similarity data of the extracted reference phonemic labels to generate a series of phonemic labels having a maximum similarity sum as a recognition result.
Claim 2
Dependent Claim
182 words
An acoustic pattern recognition apparatus for automatically recognizing continuous input speech, comprising acoustic analysis means (10) for dividing an acoustic pattern of the continous input speech at predetermined time intervals so as to produce a plurality of frame data, dictionary memory means (14) for storing reference acoustic patterns of phonemes in a selected language as reference phonemic labels, and similarity calculation means (12) for calculating similarities between the frame data and the reference acoustic patterns so as to produce a plurality of similarity data, characterized in that main processor means (16) is connected to said similarity calculation means, said main processor means prestoring at least one phonetic/ phonological data as a condition in the processing of phonemic labels, extracting, from among the reference phonemic labels to be compared with the frame data, reference phonemic labels which satisfy the condition for the phonetic/phonological data with respect to the phonemes of the input speech, and calculating similarity sums for the similarity data of the extracted reference phonemic labels to generate a series of phonemic labels having a maximum similarity sum as a recognition result.
Claim 3
Dependent Claim
42 words
The apparatus according to claim 1, characterized in that said main processor means (16) comprises: data memory means (22, 24) for prestoring, as the phonetic/phonological condition, data of at least one of speech duration and connectability for phonemes in the selected language.
Claim 4
Dependent Claim
51 words
The apparatus according to claim 2, characterized in that said data memory means comprises: first data memory means (22) for prestoring data of upper- and lower-limit values of the speech duration of phonemes in a table format; or second data memory means (24) for prestoring data of the connectability of phonemes.
Claim 5
Dependent Claim
51 words
The apparatus according to claim 2, characterized in that said data memory means comprises: first data memory means (22) for prestoring data of upper- and lower-limit values of the speech duration of phonemes in a table format; and second data memory means (24) for prestoring data of the connectability of phonemes.
Claim 6
Dependent Claim
51 words
The apparatus according to claim 2, characterized in that said main processor means (16), when similarity data for the frame data which is unsatisfied with at least one condition for the speech duration and connectability data is present, rejects that similarity data as a candidate for the maximum similarity sum calculation.
Claim 7
Dependent Claim
37 words
The apparatus according to claim 2, characterized in that said main processor means (16) comprises: a similarity data memory which is connected to said similarity calculator means and temporarily stores the similarity data for each frame data.
Claim 8
Dependent Claim
34 words
The apparatus according to claim 2, characterized in that said main processor means (16) calculates the maximum similarity sum for the similarity data of the extracted reference phonemic labels using a dynamic programming method.
Claim 9
Dependent Claim
172 words
An acoustic pattern recognition method for automatically recognizing continuous input speech, comprising the steps of dividing an acoustic pattern of the continuous input speech at predetermined time intervals so as to produce a plurality of frame data, reading out reference acoustic patterns of phonemes in a selected language prestored in a dictionary memory (14) as reference phonemic labels, and calculating similarities between the frame data and the reference acoustic patterns so as to produce a plurality of similarity data, characterized in that said method further comprises the steps of: veryfying whether reference phonemic labels which satisfy at least one condition for the phonetic/phonological data with respect to the phonemes of the input speech are present in the reference phonemic labels to be compared with the frame data, to extract the reference phonemic labels which satisfy the condition for the phonetic/phonological data; and calculating similarity sums for the similarity data of the extracted reference phonemic labels to generate a series of phonemic labels having a maximum similarity sum as a final recognition result.
Claim 10
Dependent Claim
172 words
An acoustic pattern recognition method for automatically recognizing continuous input speech, comprising the steps of dividing an acoustic pattern of the continuous input speech at predetermined time intervals so as to produce a plurality of frame data, reading out reference acoustic patterns of phonemes in a selected language prestored in a dictionary memory (14) as reference phonemic labels, and calculating similarities between the frame data and the reference acoustic patterns so as to produce a plurality of similarity data, characterized in that said method further comprises the steps of: veryfying whether reference phonemic labels which satisfy at least one condition for the phonetic/phonological data with respect to the phonemes of the input speech are present in the reference phonemic labels to be compared with the frame data, to extract the reference phonemic labels which satisfy the condition for the phonetic/phonological data; and calculating similarity sums for the similarity data of the extracted reference phonemic labels to generate a series of phonemic labels having a maximum similarity sum as a final recognition result.
Claim 11
Dependent Claim
25 words
The method according to claim 8, characterized in that the phonetic/phonological data is data representing speech duration and connectability of phonemes in the selected language.
Claim 12
Dependent Claim
45 words
The method according to claim 8, characterized in that when similarity data for the frame data which is unsatisfied with at least one condition for the speech duration and connectability data is present, that similarity data is excluded from candidates for maximum similarity sum calculation.
Claim 13
Dependent Claim
30 words
The method according to claim 8, characterized in that a maximum similarity sum for the similarity data of the extracted reference phonemic labels is calculated using a dynamic programming method.
Claim 14
Dependent Claim
20 words
The method according to claim 8, characterized in that the similarity calculation is conducted using a complex similarity calculation method.
References Cited (Backward Citations)
GB2107101A
Cited
EP0125648A1
Cited
USPTO Annuity & Maintenance Windows
Maintenance schedule is calculated for granted US utility patents.