US4520500A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Speech recognition system
Assignee: Oki Electric Industry Co Ltd
•
Filing: 1982-04-29
•
Granted: 1985-05-28
•
Expires: 2002-04-29
Abstract
Abstract This speech recognition system uses a converted-spectrum for comparison of unknown and prestored reference signals. The comparison distance is modified by a weighting factor W to emphasize differences. A straight line approximation of the short term spectrum provides a datum against which the original spectrum is differenced to provide the converted spectrum, and provide deviation signs which are used to modify an initial prestored weighting factor P to form W.
Inventors
- 👤 Kaneyoshi Mizuno
- 👤 Isamu Nose
- 👤 Hirosi Iizuka
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-11-06 12:26
CPC Classifications
G; G10; G10L; G10L15/00; G10L25/00; G10L25/78; G10L25/87
Claims Analysis
Total Claims: 8
Claim 1
Independent Claim
414 words
A speech recognition system comprising: (a) an input terminal (11) for accepting an input speech signal to be recognized, (b) a frequency analyzer (12) coupled with said input terminal (11), for providing the spectrum of the input speech signal, (c) a spectrum converter (16) coupled with the output of the frequency analyzer (12), for converting the spectrum from said frequency analyzer (12) to the difference between said spectrum and the least square fit line of said spectrum and delivering a spectrum-converted output x I (i,L), (d) a speech detector (13) coupled with the output of said frequency analyzer (12) for detecting the start point and the end point of the input speech, (e) a distance calculation means (17) for calculating the distance between the converted spectrum of input speech and the converted spectrum of each of a number of reference data, and (f) a decision circuit (18) for recognizing an input speech by selecting the minimum distance among the outputs of said distance calculation means (17), CHARACTERIZED IN THAT said distance calculation means (17) comprises: (g) an input memory (105) for storing the output x I (i,L) of said spectrum converter (16), the storing operation being started when the start point is detected by the speech detector (13) and being finished when the end point is detected by the speech detector (13), (h) a reference memory (107) for storing the converted spectrum x R (i,L) of the reference data and an original weight provided for particular portions of the reference data, (i) a distance calculator (108) for calculating the distance |x R (i,L)-x I (i,L)| between the output x I (i,L) of said input memory (105) and the converted spectrum x R (i,L) of the output of said reference memory (107), (j) weight adjusting means (208, 209, 212; or 310, 320, 330, 340, 350) for adjusting the original weight read out from the reference memory (107) according to the output x I (i,L) from the input memory (105) and the converted spectrum x R (i,L) of the reference data from the reference memory (107), and providing the weight w(i,L), (k) a multiplier (213) for providing w(i,L)x|x R (i,L)-x I (i,L)| by multiplying the output |x R (i,L)-x I (i,L)| of said distance calculator (108) and the weight w(i,L) provided by said weight adjusting means (208, 209, 212; or 310, 320, 330, 340, 350), and (l) means for accumulating the output of said multiplier (213) to provide the resultant distance to said decision circuit (18).
Claim 2
Dependent Claim
414 words
A speech recognition system comprising: (a) an input terminal (11) for accepting an input speech signal to be recognized, (b) a frequency analyzer (12) coupled with said input terminal (11), for providing the spectrum of the input speech signal, (c) a spectrum converter (16) coupled with the output of the frequency analyzer (12), for converting the spectrum from said frequency analyzer (12) to the difference between said spectrum and the least square fit line of said spectrum and delivering a spectrum-converted output x I (i,L), (d) a speech detector (13) coupled with the output of said frequency analyzer (12) for detecting the start point and the end point of the input speech, (e) a distance calculation means (17) for calculating the distance between the converted spectrum of input speech and the converted spectrum of each of a number of reference data, and (f) a decision circuit (18) for recognizing an input speech by selecting the minimum distance among the outputs of said distance calculation means (17), CHARACTERIZED IN THAT said distance calculation means (17) comprises: (g) an input memory (105) for storing the output x I (i,L) of said spectrum converter (16), the storing operation being started when the start point is detected by the speech detector (13) and being finished when the end point is detected by the speech detector (13), (h) a reference memory (107) for storing the converted spectrum x R (i,L) of the reference data and an original weight provided for particular portions of the reference data, (i) a distance calculator (108) for calculating the distance |x R (i,L)-x I (i,L)| between the output x I (i,L) of said input memory (105) and the converted spectrum x R (i,L) of the output of said reference memory (107), (j) weight adjusting means (208, 209, 212; or 310, 320, 330, 340, 350) for adjusting the original weight read out from the reference memory (107) according to the output x I (i,L) from the input memory (105) and the converted spectrum x R (i,L) of the reference data from the reference memory (107), and providing the weight w(i,L), (k) a multiplier (213) for providing w(i,L)x|x R (i,L)-x I (i,L)| by multiplying the output |x R (i,L)-x I (i,L)| of said distance calculator (108) and the weight w(i,L) provided by said weight adjusting means (208, 209, 212; or 310, 320, 330, 340, 350), and (l) means for accumulating the output of said multiplier (213) to provide the resultant distance to said decision circuit (18).
Claim 3
Dependent Claim
100 words
A speech recognition system according to claim 1, wherein the original weight obtained from the reference memory (107) is indicated by a weight indication P(i,L) which indicates an area of the reference memory (107) to be weighted and P(i,L) is 0 or 1, and said weight adjusting means provides the weight w(i,L)=1 when the weight indication P(i,L) from the reference memory is 0, and said weight adjusting means provides the weight w(i,L) equal to or larger than 1 according to the combination of the levels of the input data and the reference data when the weight indication P(i,L) is 1.
Claim 4
Dependent Claim
26 words
A speech recognition system according to claim 2, wherein the weight w(i,L) from said weight adjusting means is an integer selected from 1, 2 and 4.
Claim 5
Dependent Claim
37 words
A speech recognition system according to claim 3, wherein the weight w(i,L) from said weight adjusting means is larger than 2 when the sign of the input data is opposite from the sign of the reference data.
Claim 6
Dependent Claim
141 words
A speech recognition system according to claim 1, wherein the original weight obtained from the reference memory is indicated by weight constant WC(i,L) which is a predetermined weight for each element of the reference data stored in the reference memory (107), the weight constant WC(i,L) being either larger than 1 or smaller than 1, said weight adjusting means provides the weight w(i,L)=1 when the weight constant WC(i,L) is larger than 1 and the sign of the input data is equal to the sign of the reference data, said weight adjusting means providing the same weight as the weight constant WC(i,L) from the reference memory when the weight constant WC(i,L) is larger than 1 and the sign of the input data is opposite from the sign of the reference data, or the weight constant WC(i,L) is equal to or less than 1.
Claim 7
Dependent Claim
50 words
A speech recognition system according to claim 2, wherein said weight adjusting means includes a table memory, the addresses of the table memory being designated by the weight indication P(i,L), the level of the input data, and the level of the reference data, said table memory storing the adjusted weights.
Claim 8
Dependent Claim
162 words
A speech recognition system according to claim 5, wherein the weight adjusting means includes a sign comparator for comparing the sign of the input data with the sign of the reference data, a weight comparator for determining if the weight constant WC(i,L) is larger or smaller than 1, an AND circuit which provides an active signal when said sign comparator indicates that the sign of the input data is opposite from the sign of the reference data and said weight comparator indicates that the weight constant WC(i,L) is equal to or larger than 1, an OR circuit which provides an active output when said AND circuit provides an active output and said weight comparator indicates that the weight constant WC(i,L) is less than 1, and a switch which outputs the weight 1 when the output of said OR circuit is nonactive, and passes the weight constant WC(i,L) to the output of the switch when the output of said OR circuit is active.
References Cited (Backward Citations)
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 1988-05-27 | 1988-11-26 | 1989-05-28 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 1992-05-27 | 1992-11-26 | 1993-05-28 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 1996-05-27 | 1996-11-26 | 1997-05-28 | $7700.00 | Window Closed / Lapsed |