Patent US5151940A successfully refreshed from Google Patents.
US5151940A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Method and apparatus for extracting isolated speech word
Assignee: Fujitsu Ltd
•
Filing: 1990-12-07
•
Granted: 1992-09-29
•
Expires: 2010-12-07
Abstract
Abstract An isolated speech word is extracted from an input speech signal. First, the speech signal is divided into a high frequency band signal and a low frequency band signal. Next, the power levels of each of the high and low frequency band signals are independently compared with respective threshold levels. Finally, the front and the end of the speech word are detected from these independent comparisons.
Inventors
- 👤 Makoto Okazaki
- 👤 Koji Eto
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-10-28 23:37
CPC Classifications
G; G10; G10L; G10L15/00
Claims Analysis
Total Claims: 13
Claim 1
Independent Claim
336 words
A method for extracting an isolated speech word, having a starting point and a finishing point, from an input speech signal, said method comprising the steps of: (a) dividing the input speech signal into a high frequency band signal within a high frequency band and a low frequency band signal within a low frequency band, including the substeps of: i) obtaining the high frequency band signal from a consonant of the input speech signal; and ii) obtaining the low frequency band signal from a vowel of the input speech signal; iii) calculating an average high frequency band signal power of signal components of the high frequency band signal having predetermined frequencies within the high frequency band; iv) calculating an average low frequency band signal power of signal components of the low frequency band signal having predetermined frequencies within the low frequency band; (d) determining the starting point of the isolated speech word from the input speech signal when one of the average high frequency band signal power and the average low frequency band signal power is higher than first and second predetermined threshold levels, respectively, including the substeps of: i) comparing the average high frequency band signal power with the first threshold level; ii) comparing the average low frequency band signal power with the second threshold level; iii) detecting whether one of a first state and a second state is maintained for a predetermined term, in the first state the average high frequency band signal power is higher than the first threshold level and in the second state the average low frequency band signal power is higher than the second threshold level; and iv) identifying a starting point of the isolated speech word where one of the first and second states begins; and (e) determining the finishing point of the isolated speech word from the input speech signal when one of the average high frequency band signal power and the average low frequency band signal power is lower than third and fourth predetermined threshold levels, respectively.
Claim 2
Independent Claim
336 words
A method for extracting an isolated speech word, having a starting point and a finishing point, from an input speech signal, said method comprising the steps of: (a) dividing the input speech signal into a high frequency band signal within a high frequency band and a low frequency band signal within a low frequency band, including the substeps of: i) obtaining the high frequency band signal from a consonant of the input speech signal; and ii) obtaining the low frequency band signal from a vowel of the input speech signal; iii) calculating an average high frequency band signal power of signal components of the high frequency band signal having predetermined frequencies within the high frequency band; iv) calculating an average low frequency band signal power of signal components of the low frequency band signal having predetermined frequencies within the low frequency band; (d) determining the starting point of the isolated speech word from the input speech signal when one of the average high frequency band signal power and the average low frequency band signal power is higher than first and second predetermined threshold levels, respectively, including the substeps of: i) comparing the average high frequency band signal power with the first threshold level; ii) comparing the average low frequency band signal power with the second threshold level; iii) detecting whether one of a first state and a second state is maintained for a predetermined term, in the first state the average high frequency band signal power is higher than the first threshold level and in the second state the average low frequency band signal power is higher than the second threshold level; and iv) identifying a starting point of the isolated speech word where one of the first and second states begins; and (e) determining the finishing point of the isolated speech word from the input speech signal when one of the average high frequency band signal power and the average low frequency band signal power is lower than third and fourth predetermined threshold levels, respectively.
Claim 3
Dependent Claim
118 words
A method as set forth in claim 1, wherein said step (e) includes the steps of: (i) comparing the average high frequency band signal power with the third threshold level; (ii) comparing the average low frequency band signal power with the fourth threshold level; (iii) detecting whether one of a third state and a fourth state is maintained for a predetermined term, in the third state the average high frequency band signal power is lower than the third threshold level and in the fourth state the average low frequency band signal is lower than the fourth threshold level; and (iv) identifying the finishing point of the isolated speech word where one of the third and fourth states begins.
Claim 4
Dependent Claim
22 words
A method as set forth in claim 2, wherein said steps (b), (c), (d) and (e) are carried out by digital processing.
Claim 5
Independent Claim
314 words
A method for extracting an isolated speech word from an input speech signal, comprising: first means for receiving the input speech signal and for detecting an average consonant power therefrom, said first means includes at least a speech analyzer for receiving the input speech signal and for producing a consonant signal and a vowel signal separately therefrom; and a first calculating unit, operatively connected to said speech analyzer, for calculating an average consonant power based on the consonant signal; second means for receiving the input speech signal and for detecting an average vowel power therefrom, said second means includes at least said speech analyzer, commonly used with said first means; and a second calculating unit, operatively connected to said speech analyzer, for calculating the average vowel power based on the vowel signal; and third means for extracting the isolated speech word from the input speech signal in dependence upon the average consonant power and the average vowel power, said third means including: a first detecting unit, operatively connected to receive the average consonant power from said first calculating unit, for detecting one of a starting point and a finishing point of the consonant signal; a second detecting unit, operatively connected to receive the average vowel power from said second calculating unit, for detecting one of a starting point and a finishing point of the vowel signal; a discriminating unit, operatively connected to said first and second detecting units, for discriminating one of the starting point of the consonant signal and the vowel signal and for discriminating one of the finishing point of the consonant signal and the vowel signal; and an extracting unit, operatively connected to said speech analyzer and said discriminating unit, for receiving the input speech signal and for extracting the isolated speech word using the starting point and the finishing point of one of the consonant signal and the vowel signal.
Claim 6
Independent Claim
314 words
A method for extracting an isolated speech word from an input speech signal, comprising: first means for receiving the input speech signal and for detecting an average consonant power therefrom, said first means includes at least a speech analyzer for receiving the input speech signal and for producing a consonant signal and a vowel signal separately therefrom; and a first calculating unit, operatively connected to said speech analyzer, for calculating an average consonant power based on the consonant signal; second means for receiving the input speech signal and for detecting an average vowel power therefrom, said second means includes at least said speech analyzer, commonly used with said first means; and a second calculating unit, operatively connected to said speech analyzer, for calculating the average vowel power based on the vowel signal; and third means for extracting the isolated speech word from the input speech signal in dependence upon the average consonant power and the average vowel power, said third means including: a first detecting unit, operatively connected to receive the average consonant power from said first calculating unit, for detecting one of a starting point and a finishing point of the consonant signal; a second detecting unit, operatively connected to receive the average vowel power from said second calculating unit, for detecting one of a starting point and a finishing point of the vowel signal; a discriminating unit, operatively connected to said first and second detecting units, for discriminating one of the starting point of the consonant signal and the vowel signal and for discriminating one of the finishing point of the consonant signal and the vowel signal; and an extracting unit, operatively connected to said speech analyzer and said discriminating unit, for receiving the input speech signal and for extracting the isolated speech word using the starting point and the finishing point of one of the consonant signal and the vowel signal.
Claim 7
Dependent Claim
143 words
An apparatus as set forth in claim 4, wherein said first detecting unit detects the starting point of the consonant signal under a condition where the average consonant power is higher than a first predetermined threshold level and continues for a first predetermined term, and detects the finishing point of the consonant signal under a condition where the average consonant power is lower than a third predetermined level and continues for a third predetermined term, and wherein said second detecting unit detects the starting point of the vowel signal under a condition where the average vowel power is higher than a second predetermined threshold level and continues for a second predetermined term, and detects the finishing point of the vowel signal under a condition where the average vowel power is lower than a fourth predetermined level and continues for a fourth predetermined term.
Claim 8
Dependent Claim
192 words
An apparatus as set forth in claim 4, wherein said speech analyzer comprises: a filter band having a high frequency filtering group for the detection of the consonant signal and a low frequency filtering group for the detection of the vowel signal; a rectifier bank having a high frequency side rectifier circuit group connected to the high frequency filtering group and a low frequency side rectifier circuit group connected to the low frequency filtering group; a sampling and a holding bank having a high frequency side sampling and holding circuit group connected to the high frequency side rectifier circuit group and a low frequency side sampling and holding circuit group connected to the low frequency side rectifier circuit group; an analog multiplexer, operatively connected to the high frequency side and low frequency side sampling and holding circuit groups, for producing data in the form of a frame consisting of a plurality of analog channels indicating the consonant analog power detected at predetermined high frequencies and the vowel analog power detected at predetermined low frequencies; and an analog/digital converter, operatively connected to said analog multiplexer, for transforming the analog channels into digital data.
Claim 9
Dependent Claim
51 words
An apparatus as set forth in claim 4, wherein said first and second detecting units, said discriminating unit and said extracting unit are all realized by a microprocessor cooperating with a memory, and the microprocessor executes programs to determine the starting position and the finishing position of the input speech signal.
Claim 10
Dependent Claim
24 words
An apparatus as set forth in claim 5, wherein the first and second predetermined terms are smaller than the third and fourth predetermined terms.
Claim 11
Dependent Claim
26 words
An apparatus as set forth in claim 8, wherein the first and second predetermined terms are equal, and the third and fourth predetermined terms are equal.
Claim 12
Independent Claim
335 words
A method for extracting an isolated speech word, having a starting point and a finishing point, from an input speech signal, said method comprising the steps of: (a) dividing the input speech signal into a high frequency band signal within a high frequency band and a low frequency band signal within a low frequency band, including the substeps of: i) obtaining the high frequency band signal from a consonant of the input speech signal; and ii) obtaining the low frequency band signal from a vowel of the input speech signal; (b) calculating an average high frequency band signal power of signal components of the high frequency band signal having predetermined frequencies within the high frequency band; (c) calculating an average low frequency band signal power of signal components of the low frequency band signal having predetermined frequencies within the low frequency band; (d) determining the starting point of the isolated speech word from the input speech signal when one of the average high frequency band signal power and the average low frequency band signal power is higher than first and second predetermined threshold levels, respectively; and (e) determining the finishing point of the isolated speech word from the input speech signal when one of the average high frequency band signal power and the average low frequency band signal power is lower than third and fourth predetermined threshold levels, respectively, including the substeps of: i) comparing the average high frequency band signal power with the third threshold level; ii) comparing the average low frequency band signal power with the fourth threshold level; iii) detecting whether one of a third state and a fourth state is maintained for a predetermined term, in the third state the average high frequency band signal power is lower than the third threshold level and in the fourth state the average low frequency band signal is lower than the fourth threshold level; and iv) identifying the finishing point of the isolated speech word where one of the third and fourth states begin.
Claim 13
Independent Claim
335 words
A method for extracting an isolated speech word, having a starting point and a finishing point, from an input speech signal, said method comprising the steps of: (a) dividing the input speech signal into a high frequency band signal within a high frequency band and a low frequency band signal within a low frequency band, including the substeps of: i) obtaining the high frequency band signal from a consonant of the input speech signal; and ii) obtaining the low frequency band signal from a vowel of the input speech signal; (b) calculating an average high frequency band signal power of signal components of the high frequency band signal having predetermined frequencies within the high frequency band; (c) calculating an average low frequency band signal power of signal components of the low frequency band signal having predetermined frequencies within the low frequency band; (d) determining the starting point of the isolated speech word from the input speech signal when one of the average high frequency band signal power and the average low frequency band signal power is higher than first and second predetermined threshold levels, respectively; and (e) determining the finishing point of the isolated speech word from the input speech signal when one of the average high frequency band signal power and the average low frequency band signal power is lower than third and fourth predetermined threshold levels, respectively, including the substeps of: i) comparing the average high frequency band signal power with the third threshold level; ii) comparing the average low frequency band signal power with the fourth threshold level; iii) detecting whether one of a third state and a fourth state is maintained for a predetermined term, in the third state the average high frequency band signal power is lower than the third threshold level and in the fourth state the average low frequency band signal is lower than the fourth threshold level; and iv) identifying the finishing point of the isolated speech word where one of the third and fourth states begin.
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) | 1995-09-29 | 1996-03-30 | 1996-09-29 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 1999-09-29 | 2000-03-30 | 2000-09-29 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2003-09-29 | 2004-03-30 | 2004-09-29 | $7700.00 | Window Closed / Lapsed |