WO1987006384A1 - Method of collating voice pattern in voice recognizing apparatus - Google Patents
Method of collating voice pattern in voice recognizing apparatus Download PDFInfo
- Publication number
- WO1987006384A1 WO1987006384A1 PCT/JP1986/000191 JP8600191W WO8706384A1 WO 1987006384 A1 WO1987006384 A1 WO 1987006384A1 JP 8600191 W JP8600191 W JP 8600191W WO 8706384 A1 WO8706384 A1 WO 8706384A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- selection
- voice
- result
- preliminary
- main
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims description 18
- 230000000694 effects Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/78—Detection of presence or absence of voice signals
- G10L25/87—Detection of discrete points within a voice signal
-
- C—CHEMISTRY; METALLURGY
- C08—ORGANIC MACROMOLECULAR COMPOUNDS; THEIR PREPARATION OR CHEMICAL WORKING-UP; COMPOSITIONS BASED THEREON
- C08G—MACROMOLECULAR COMPOUNDS OBTAINED OTHERWISE THAN BY REACTIONS ONLY INVOLVING UNSATURATED CARBON-TO-CARBON BONDS
- C08G2150/00—Compositions for coatings
Definitions
- Speech pattern matching method for speech recognition device
- the present invention relates to a pattern matching method in a voice recognition device.
- the present invention has been made to solve the above-mentioned drawbacks of the prior art, and has an object, in particular, to perform recognition processing at a higher speed in a speech recognition apparatus having two functions of preliminary selection and main selection. It was done as.
- the present invention performs preliminary selection using a certain range of data at the beginning of an unknown voice, and performs preliminary selection based on information at the beginning of a speech recognition device that performs this selection after utterance is completed. If the audio lock signal changes during the operation, and a silent section is reached, the preliminary selection is interrupted and the process proceeds to the main selection process.
- the feature is that when the utterance is finished as it is, the result is output, and when returning to the voice section again, the main selection is interrupted or the result is discarded and the preliminary selection is performed.
- 1 to 3 are time charts for explaining an embodiment of the voice pattern matching method according to the present invention.
- Fig. 1 is a time chart for explaining the processing when no silence section is included in the speech section.
- (A) is a one-word pulse
- (b) is a speech section pulse
- (o) is a pulse.
- the speech section pulse (b) is a pulse that detects the speech power signal at a certain threshold
- the one-word pulse (a) is a pause section after the speech section pulse ends (generally 200 ⁇ 35 O ns).
- A is the initial data capturing section
- B is the preliminary selection section
- C is the main selection section
- D is the result output section, and performs preliminary selection based on the audio data of the first n frames.
- the top words from the target word in the dictionary up to a certain number of words are extracted, and the result is subjected to the main selection processing and the result is output.
- the pre-selection time is not sufficient. This hindered the selection and prevented the recognition processing time from being shortened.
- Fig. 2 is a time chart for explaining the processing when a silent section is included. First, a certain range of audio data from the word
- the preliminary selection B Immediately enters the preliminary selection B, and immediately enters the preliminary selection B, and if the voiced end edge a of the voice section pulse occurs while performing the preliminary selection 5, the preliminary selection is immediately interrupted and the preliminary selection is performed. After all of the intermediate results and the word currently being processed are stored, full selection C of the section of X is performed. In this secrecy, firstly, the result of the preliminary selection up to the middle is used, and
- the main selection is performed based on the results of the 10 selections, and after that, the main selection is performed for all words and the final result (the one with the highest similarity) is obtained. If the start edge b of the next sound section comes during the main selection process, the process of the main selection C is interrupted, and the result of the main selection is invalidated. Also, this selection has already been completed and
- step 20 When step 20 is completed, the result D of this selection is output at that point.
- Be will perform only the main selection based on that, and will make the main selection at each start edge of the silent section, and the one-word pulse will be generated. Wait for completion, and wait for the result output. By doing so, Even when the word length is short, the most efficient processing can be performed, and the optimal speed reduction processing by preliminary selection becomes possible.
- FIG. 3 is a time chart showing an example of processing in a case where the preliminary selection has been completed in the case shown in FIG. 2 even though the sound period has continued for a relatively long time.
- first preliminary selection is ended, carried Ri by its pre-selection was based on the second performs preliminary selection B 2 by (eg prefix n full record over arm of the first preliminary selection B i, of the second preselection B 2 is carried out in m frames comprising n Yo Li Univ from prefix), similarly to the above sound, the selection process time from word endings be performed in conjunction with the present selected to silence can further shorten , More words can be handled, and-this selection processing time can be further shortened.
- the optimum high-speed processing can be performed for any word length. Furthermore, by using the present invention for pattern matching processing, optimal high-speed processing can be performed for any word length, the processing time from the end can be shortened, and the recognition rate can be improved. Even if the number of words in the book is increased, the effect on the processing speed can be reduced.
- the speech pattern identification method according to the present invention can be applied to speech recognition.
Landscapes
- Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Telephonic Communication Services (AREA)
- Document Processing Apparatus (AREA)
Description
Claims
Priority Applications (5)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
DE19863690416 DE3690416T1 (ja) | 1986-04-16 | 1986-04-16 | |
PCT/JP1986/000191 WO1987006384A1 (en) | 1986-04-16 | 1986-04-16 | Method of collating voice pattern in voice recognizing apparatus |
EP86902496A EP0266423B1 (en) | 1986-04-16 | 1986-04-16 | Method of collating voice pattern in voice recognizing apparatus |
GB8629794A GB2194661B (en) | 1986-04-16 | 1986-04-16 | Voice pattern matching method of a voice recognition unit |
DE19863690416 DE3690416C2 (de) | 1986-04-16 | 1986-04-16 | Sprach- und Stimmenerkennungsverfahren |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/JP1986/000191 WO1987006384A1 (en) | 1986-04-16 | 1986-04-16 | Method of collating voice pattern in voice recognizing apparatus |
Publications (1)
Publication Number | Publication Date |
---|---|
WO1987006384A1 true WO1987006384A1 (en) | 1987-10-22 |
Family
ID=13874427
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP1986/000191 WO1987006384A1 (en) | 1986-04-16 | 1986-04-16 | Method of collating voice pattern in voice recognizing apparatus |
Country Status (4)
Country | Link |
---|---|
EP (1) | EP0266423B1 (ja) |
DE (2) | DE3690416T1 (ja) |
GB (1) | GB2194661B (ja) |
WO (1) | WO1987006384A1 (ja) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0535146A1 (en) * | 1990-06-22 | 1993-04-07 | Dragon Systems Inc | CONTINUOUS SPEECH PROCESSING SYSTEM. |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH01167898A (ja) * | 1987-12-04 | 1989-07-03 | Internatl Business Mach Corp <Ibm> | 音声認識装置 |
JP3045510B2 (ja) * | 1989-12-06 | 2000-05-29 | 富士通株式会社 | 音声認識処理装置 |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS576900A (en) * | 1980-06-16 | 1982-01-13 | Nippon Telegraph & Telephone | Voice pattern recognition system |
JPS6195397A (ja) * | 1984-10-17 | 1986-05-14 | 株式会社リコー | 音声パターン照合方法 |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4370521A (en) * | 1980-12-19 | 1983-01-25 | Bell Telephone Laboratories, Incorporated | Endpoint detector |
-
1986
- 1986-04-16 DE DE19863690416 patent/DE3690416T1/de not_active Withdrawn
- 1986-04-16 GB GB8629794A patent/GB2194661B/en not_active Expired
- 1986-04-16 WO PCT/JP1986/000191 patent/WO1987006384A1/ja active IP Right Grant
- 1986-04-16 DE DE19863690416 patent/DE3690416C2/de not_active Expired - Lifetime
- 1986-04-16 EP EP86902496A patent/EP0266423B1/en not_active Expired - Lifetime
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS576900A (en) * | 1980-06-16 | 1982-01-13 | Nippon Telegraph & Telephone | Voice pattern recognition system |
JPS6195397A (ja) * | 1984-10-17 | 1986-05-14 | 株式会社リコー | 音声パターン照合方法 |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0535146A1 (en) * | 1990-06-22 | 1993-04-07 | Dragon Systems Inc | CONTINUOUS SPEECH PROCESSING SYSTEM. |
EP0535146A4 (ja) * | 1990-06-22 | 1995-01-04 | Dragon Systems Inc |
Also Published As
Publication number | Publication date |
---|---|
EP0266423A4 (en) | 1989-03-29 |
EP0266423B1 (en) | 1994-03-09 |
EP0266423A1 (en) | 1988-05-11 |
GB2194661A (en) | 1988-03-09 |
DE3690416C2 (de) | 1990-08-16 |
GB2194661B (en) | 1989-12-06 |
DE3690416T1 (ja) | 1988-03-10 |
GB8629794D0 (en) | 1987-01-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP3789246B2 (ja) | 音声区間検出装置、音声区間検出方法、音声認識装置、音声認識方法および記録媒体 | |
CN102971787A (zh) | 录音结束点检测方法及系统 | |
WO1987006384A1 (en) | Method of collating voice pattern in voice recognizing apparatus | |
JPS6123560B2 (ja) | ||
JPS6195397A (ja) | 音声パターン照合方法 | |
JPS6344699A (ja) | 音声認識装置 | |
JPS5887599A (ja) | 音声単語認識装置 | |
JP2891259B2 (ja) | 音声区間検出装置 | |
JPH0567036B2 (ja) | ||
JPH05303391A (ja) | 音声認識装置 | |
JPH01158499A (ja) | 定常雑音除去方式 | |
JPH0262879B2 (ja) | ||
JPS60233700A (ja) | 音声パタ−ン照合方法 | |
JPS59111697A (ja) | 音声認識方式 | |
JPS61260299A (ja) | 音声認識装置 | |
JPS61252595A (ja) | 音声認識処理方式 | |
JP2712586B2 (ja) | 単語音声認識装置用パターンマッチング方式 | |
JPS6195399A (ja) | 音声パターン照合方法 | |
JPS61254994A (ja) | 連続単語認識装置 | |
JPS6135498A (ja) | 音声認識装置 | |
JPS62116999A (ja) | 音節単位音声認識装置 | |
JPS61256397A (ja) | 音声認識装置 | |
JPS62293299A (ja) | 音声認識方法 | |
JPS6287993A (ja) | 音声認識装置 | |
JPS6027000A (ja) | パタンマツチング方法 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): DE GB US |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): FR |
|
WWE | Wipo information: entry into national phase |
Ref document number: 1986902496 Country of ref document: EP |
|
RET | De translation (de og part 6b) |
Ref document number: 3690416 Country of ref document: DE Date of ref document: 19880310 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 3690416 Country of ref document: DE |
|
WWP | Wipo information: published in national office |
Ref document number: 1986902496 Country of ref document: EP |
|
WWG | Wipo information: grant in national office |
Ref document number: 1986902496 Country of ref document: EP |