EP0086589A1
Expired / Lapsed
EP
⏱️ 0.0 years remaining
Speech recognition system
Assignee: Toshiba Corp
•
Filing: 1983-01-31
•
Granted: —
•
Expires: 2003-01-31
Abstract
Abstract An acoustic signal processing circuit (12) extracts input speech pattern data and subsidiary feature data from an input speech signal. The input speech pattern data comprise frequency spectra, whereas the subsidiary feature data comprise phoneme and acoustic features. These data are then stored in a data buffer memory (13). The similarity measures between the input speech pattern data stored in the data buffer memory (13) and reference speech pattern data stored in a dictionary memory (16) are computed by a similarity computation circuit (15). When the largest similarity measure exceeds a first threshold value and when the difference between the largest similiarity measure and the second largest measure exceeds a second threshold value, category data of the reference pattern which gives the largest similarity measure is produced by a control circuit (14) to correspond to an input speech. When recognition cannot be performed, the categories of the reference speech patterns which respectively give the largest to mth similiarity measures are respectively compared with the subsidiary feature data. In this manner, subsidiary feature recognition of the input voice is performed by a subsidiary feature recognition section (17).
Inventors
- 👤 Sadakazu Watanabe
- 👤 Hidenori Shinoda
- 👤 Tsuneo Nitta
- 👤 Yoichi Takebayashi
- 👤 Shouichi Hirai
- 👤 Tomio Sakata
- 👤 Kensuke Uehara
- 👤 Yasuo Takahashi
- 👤 Haruo C/O Prof. P.H. Winston The Artificial Asada
Bibliographic Details
Application Number
—
Jurisdiction
EP
Data Source
google_patents
Last Synchronized
2026-10-31 22:00
CPC Classifications
G; G10; G10L; G10L15/00
Claims Analysis
Total Claims: 12
Claim 1
Independent Claim
228 words
A speech recognition system characterized by comprising: means (12) for extracting input speech pattern data which consists of time series of first feature parameter data of an input speech from an input speech signal; a buffer memory (13) for storing the input speech pattern data; a reference pattern memory (16) for storing a plurality of reference speech pattern data; a similarity computation circuit (15) for computing a similarity measure between the input speech pattern data stored in said buffer memory (13) and said plurality of reference speech pattern data stored in said reference pattern memory (16); comparing means (14, 17) for determining whether or not a category having a reference pattern which gives a largest similarity measure corresponds to the input speech in accordance with a difference between the largest similarity measure and a second largest similarity measure which are obtained by said similarity computation circuit (15), for extracting m categories of said plurality of reference patterns which give the largest to mth largest similarity measures when the category is regarded not to correspond to the input speech, and for comparing the m categories with the input speech by using second feature parameter data; and memory means (13) for storing the second feature parameter data, whereby the category of the reference pattern corresponding to the input speech is determined by comparison results by said comparing means (14, 17).
Claim 2
Dependent Claim
228 words
A speech recognition system characterized by comprising: means (12) for extracting input speech pattern data which consists of time series of first feature parameter data of an input speech from an input speech signal; a buffer memory (13) for storing the input speech pattern data; a reference pattern memory (16) for storing a plurality of reference speech pattern data; a similarity computation circuit (15) for computing a similarity measure between the input speech pattern data stored in said buffer memory (13) and said plurality of reference speech pattern data stored in said reference pattern memory (16); comparing means (14, 17) for determining whether or not a category having a reference pattern which gives a largest similarity measure corresponds to the input speech in accordance with a difference between the largest similarity measure and a second largest similarity measure which are obtained by said similarity computation circuit (15), for extracting m categories of said plurality of reference patterns which give the largest to mth largest similarity measures when the category is regarded not to correspond to the input speech, and for comparing the m categories with the input speech by using second feature parameter data; and memory means (13) for storing the second feature parameter data, whereby the category of the reference pattern corresponding to the input speech is determined by comparison results by said comparing means (14, 17).
Claim 3
Dependent Claim
27 words
A system according to claim 1, characterized in that said input speech pattern extracting-means (12) includes means for obtaining first frequency spectrum data of the input speech.
Claim 4
Dependent Claim
26 words
A system according to claim 2, characterized in that said first frequency spectrum data obtaining means includes a digital band-pass filter which performs time sharing operation.
Claim 5
Dependent Claim
29 words
A system according to claim 2, characterized in that said input speech pattern data extracting means (12) includes means for obtaining second frequency spectrum data of the input speech.
Claim 6
Dependent Claim
54 words
A system according to claim 4, characterized in that said second frequency spectrum data obtaining means includes a digital filter comprising: a coefficient memory (29) for prestoring coefficient data to set filter characteristics corresponding to the first and second frequency spectrum data; and a micro-programmable controller (22) for controlling operation of said digital filter.
Claim 7
Dependent Claim
92 words
A system according to claim 1, characterized in that said similarity computation circuit (15) includes: a micro-programmable controller (62) for reading out the input speech pattern data stored in said buffer memory (13) and the reference speech pattern data stored in said reference pattern memory (16); a multiplier (69) for multiplying the input speech pattern data and the reference speech pattern data; operation circuits (71, 72) for sequentially accumulating outputs from said multiplier (69); and a register (73) for storing as similarity measure data an output from said operation circuits (71, 72).
Claim 8
Dependent Claim
37 words
A system according to claim 6, characterized in that said similarity computation circuit (15) further has sorting means (77, 78, 81) for arranging m similarity measure data stored in said register (73) in accordance with magnitudes thereof.
Claim 9
Dependent Claim
89 words
A system according to claim 4, characterized in that said input speech pattern data extracting means (12) has phoneme feature data extracting means (27, 29, 30, 32) for extracting phoneme feature data such as vowel and nasal sounds from the input speech signal, and said comparing means (14, 17) has acoustic feature data extracting means (50, 52, 53) for extracting acoustic feature data from the second frequency spectrum data and performs speech recognition using as the second feature parameter data the phoneme feature data and the acoustic feature data.
Claim 10
Dependent Claim
80 words
A system according to claim 1, characterized in that said comparing means (14, 17) has a resistor (53) for storing first and second threshold values, and a controller (50) for determining that the category of the reference speech pattern which gives the largest similarity measure corresponds to a finally recognized speech when the largest similarity measure exceeds the first threshold value and the difference between the' largest similarity measure and the second largest similarity measure exceeds the second threshold value.
Claim 11
Independent Claim
170 words
A method for recognizing speech comprising the steps of: extracting input speech pattern data which consists of time series of first feature parameter data of an input speech from an input speech signal; storing the input speech pattern data; storing a plurality of reference speech pattern data; computing a similarity measure between the input speech pattern data and said plurality of. reference speech pattern data; determining whether or not a category having a reference pattern which gives a largest similarity measure corresponds to the input speech in accordance with a difference between the largest similarity measure and a second largest similarity measure; extracting m categories of said plurality of reference patterns which give the largest to-mth largest similarity measures when the category is regarded not to correspond to the input speech; - comparing the m categories with the input speech by using second feature parameter data; and storing the second feature parameter data, whereby the category of the reference pattern corresponding to the input speech is determined by comparison results.
Claim 12
Independent Claim
170 words
A method for recognizing speech comprising the steps of: extracting input speech pattern data which consists of time series of first feature parameter data of an input speech from an input speech signal; storing the input speech pattern data; storing a plurality of reference speech pattern data; computing a similarity measure between the input speech pattern data and said plurality of. reference speech pattern data; determining whether or not a category having a reference pattern which gives a largest similarity measure corresponds to the input speech in accordance with a difference between the largest similarity measure and a second largest similarity measure; extracting m categories of said plurality of reference patterns which give the largest to-mth largest similarity measures when the category is regarded not to correspond to the input speech; - comparing the m categories with the input speech by using second feature parameter data; and storing the second feature parameter data, whereby the category of the reference pattern corresponding to the input speech is determined by comparison results.
References Cited (Backward Citations)
USPTO Annuity & Maintenance Windows
Maintenance schedule is calculated for granted US utility patents.