公开(公告)号：JP2002091756A

公开(公告)日：2002-03-29

申请号：JP2001178004

申请日：2001-06-13

Applicant: IBM

Inventor： GONG QUING ,  LEWIS JAMES R ,  VANBUSKIRK RONALD E ,  WANG HUIFANG

IPC: G06F3/16 ,  G10L15/00 ,  G10L15/22 ,  H04H20/88 ,  H04M3/50 ,  H04H5/00

Abstract: PROBLEM TO BE SOLVED: To enable a user to select one of plural acoustic information sources in voice when acoustic information from at least two acoustic information sources simultaneously exists, and next to acoustically discriminate the acoustic information source selected by the user from the other acoustic information source by resetting at least one of plural acoustic information sources. SOLUTION: In the method for simultaneously providing a large number of information sources, the acoustic information is simultaneously provided from at least two acoustic information sources and the user voice selection of at least one acoustic information source is permitted. Besides, at least one acoustic information source is reset. Further, at least one acoustic information source selected by the user is acoustically identified from the other acoustic information source by such resetting.

公开(公告)号：JP2000259170A

公开(公告)日：2000-09-22

申请号：JP2000027657

申请日：2000-02-04

Applicant: IBM

Inventor： JAMES R LEWIS ,  WANG HUIFANG ,  VAN BUSKIRK RON ,  ORTEGA KERRY A

IPC: G06F3/16 ,  G10L15/00 ,  G10L15/06

Abstract: PROBLEM TO BE SOLVED: To enable a person disabled to read or device having no display to perform voice recognition and registration by instructing a user in voice so as to read a text phrase reproduced in voice. SOLUTION: The user is instructed to keep silence while reproducing a phrase and to read respective phrases after the end of reproduction of the respective phrases and a present phrase is reproduced (S26). The present phrase is reproduced in a voice 2. When the present phrase is completely reproduced, the read of that reproduced phrase due to the user is waited. When the phrase read by the user can be accepted (S44), the opportunity of selection of whether registration is to be performed at that time point or registration is to be suspended can be applied to the user as well. Thus, a person disabled to read, user disabled to read a text or user defining another language as the first language can perform registration to a voice recognition system.

公开(公告)号：JP2002073084A

公开(公告)日：2002-03-12

申请号：JP2001178028

申请日：2001-06-13

Applicant: IBM

Inventor： BADT DANIEL E ,  GUASTI PETER J ,  HANSON GARY R ,  NASSIFF AMADO ,  RODRIGUEZ EDWIN A ,  RUBACK HARVEY R ,  SMITH CARL A ,  VANBUSKIRK RONALD E ,  WANG HUIFANG ,  STEVEN G WOODWARD

IPC: G06F12/08 ,  G06F3/16 ,  G06F12/12 ,  G10L15/00 ,  G10L15/06 ,  G10L15/18 ,  G10L15/22 ,  G10L15/28

Abstract: PROBLEM TO BE SOLVED: To provide a method and a system for improving the re-readout of voice data in a computer voice system. SOLUTION: When the voice data can be compared with a prescribed voice cache entry reference/emission standard and the voice data satisfies one of the prescribed entry references, the additional purge of at least one item can be performed to a voice cache. The item corresponds to the voice data. The entry reference can include voice data used frequently, voice data used recently, and important voice data. Similarly, the emission standard can include voice data having the minimum frequency of use relating to each item in the voice caches, voice data unused for the longest time relating to each item in the voice caches, and voice data having the minimum level of importance relating to each item in the voice caches.

公开(公告)号：JP2001034293A

公开(公告)日：2001-02-09

申请号：JP2000188566

申请日：2000-06-23

Applicant: IBM

Inventor： ORTEGA KERRY A ,  EGGER HANS ,  ARTHUR KELLER ,  RONALD E VANBASUKAAKU ,  WANG HUIFANG ,  JAMES R LOUIS

IPC: G10L15/28 ,  G06F3/16 ,  G10L15/00 ,  G10L15/08 ,  G10L15/22

Abstract: PROBLEM TO BE SOLVED: To provide a function and information for changing a system parameter or the action of a user easily in order to enhance the recognizing accuracy of a transfer system. SOLUTION: An input voice is received by a system to be transferred (a step 204). The system monitors the accuracy of the transferred voice when it is transferred (a step 205). Moreover, the system decides whether the accuracy of the transferred voice is sufficient or not (a step 210) and when it is not sufficient, the system starts up a voice recognition improving tool automatically (a step 214) and alarms that the tool has been started up to the user (a step 212). The type of a recognition problem is identified by the user or is identified automatically by the system (a step 216) and the system supplies a possible resolving step for enabling the user to adjust the system parameter of for enabling to correct the action of the user in order to reduce the recognition problem (a step 218).

公开(公告)号：DE60100090T2

公开(公告)日：2003-09-25

申请号：DE60100090

申请日：2001-06-12

Applicant: IBM

Inventor： BADT DANIEL E ,  GUASTI PETER J ,  HANSON GARY R ,  NASSIFF AMADO ,  RODRIGUEZ EDWIN A ,  RUBACK HARVEY R ,  SMITH CARL A ,  VANBUSKIRK RONALD E ,  WANG HUIFANG

IPC: G06F12/08 ,  G06F3/16 ,  G06F12/12 ,  G10L15/00 ,  G10L15/06 ,  G10L15/18 ,  G10L15/22 ,  G10L15/26

Abstract: A method and a system for improving recall of speech data in a computer speech system can include a plurality of speech cache management steps including providing a speech cache, receiving a speech system input and identifying a speech event in the received speech system input, the speech event comprising speech data. Subsequently, the speech data can be compared to pre-determined speech cache entry criteria; and, if the speech data meets one of the pre-determined entry criteria, at least one entry can be added to the speech cache, the at least one entry corresponding to the speech data. Additionally, the speech data can be compared to pre-determined speech cache exit criteria; and, if the speech data meets one of the pre-determined exit criteria, at least one entry can be purged from the speech cache, the at least one entry corresponding to the speech data. The entry criteria can include frequently used speech data, recently used speech data, and important speech data. Similarly, the exit criteria can include least frequently used speech data associated with each entry in the speech cache, least recently used speech data associated with each entry in the speech cache and least important speed data associated with each entry in the speech cache.

公开(公告)号：CA2345434A1

公开(公告)日：2001-12-15

申请号：CA2345434

申请日：2001-04-27

Applicant: IBM

Inventor： LEWIS JAMES R ,  VANBUSKIRK RONALD E ,  GONG QING ,  WANG HUIFANG

IPC: G06F3/16 ,  G10L15/00 ,  G10L15/22 ,  H04H20/88 ,  H04M3/50

Abstract: A method for concurrent presentation of multiple audio information sources. In the method, audio information from at least two audio information sources is concurrentl y presented, and a user speech selection of one of the audio information sources is accepted. At lea st one of the audio information sources can then be reconfigured. The reconfiguration audibly distinguishes the user selected audio information source from other audio information sources.

公开(公告)号：DE60023398T2

公开(公告)日：2006-07-06

申请号：DE60023398

申请日：2000-05-24

Applicant: IBM

Inventor： BALLARD BARBARA ELAINE ,  LEWIS JAMES R ,  NASSIFF AMADO ,  ORTEGA KERRY A ,  VANBUSKIRK RONALD E ,  WANG HUIFANG

IPC: G10L15/18 ,  G10L15/22 ,  G10L15/26

Abstract: A method and system for improving the speech command recognition accuracy of a computer speech recognition system uses event-based constraints to recognize a spoken command. The constraints are system states and events, which include system activities, active applications, prior commands and an event queue. The method and system is performed by monitoring events and states of the computer system and receiving a processed command corresponding to the spoken command. The processed command is statistically analyzed in light of the system events and states as well as according to an acoustic model. The system then identifies a recognized command corresponding to the spoken command.

公开(公告)号：AT231643T

公开(公告)日：2003-02-15

申请号：AT01000205

申请日：2001-06-12

Applicant: IBM

Inventor： BADT DANIEL E ,  GUASTI PETER J ,  HANSON GARY R ,  NASSIFF AMADO ,  RODRIGUEZ EDWIN A ,  RUBACK HARVEY R ,  SMITH CARL A ,  VANBUSKIRK RONALD E ,  WANG HUIFANG

IPC: G06F12/08 ,  G06F3/16 ,  G06F12/12 ,  G10L15/00 ,  G10L15/06 ,  G10L15/18 ,  G10L15/22 ,  G10L15/26

Abstract: A method and a system for improving recall of speech data in a computer speech system can include a plurality of speech cache management steps including providing a speech cache, receiving a speech system input and identifying a speech event in the received speech system input, the speech event comprising speech data. Subsequently, the speech data can be compared to pre-determined speech cache entry criteria; and, if the speech data meets one of the pre-determined entry criteria, at least one entry can be added to the speech cache, the at least one entry corresponding to the speech data. Additionally, the speech data can be compared to pre-determined speech cache exit criteria; and, if the speech data meets one of the pre-determined exit criteria, at least one entry can be purged from the speech cache, the at least one entry corresponding to the speech data. The entry criteria can include frequently used speech data, recently used speech data, and important speech data. Similarly, the exit criteria can include least frequently used speech data associated with each entry in the speech cache, least recently used speech data associated with each entry in the speech cache and least important speed data associated with each entry in the speech cache.

公开(公告)号：CA2303718A1

公开(公告)日：2000-11-29

申请号：CA2303718

申请日：2000-04-05

Applicant: IBM

Inventor： BALLARD BARBARA ELAINE ,  WANG HUIFANG ,  VANBUSKIRK RONALD E ,  ORTEGA KERRY A ,  NASSIFF AMADO ,  LEWIS JAMES R

IPC: G10L15/18 ,  G10L15/22 ,  G10L15/26

Abstract: A method and system for improving the speech command recognition accuracy of a computer speech recognition system uses event-based constraints to recognize a spoken command. The constraints are system states and events, which include system activities, active applications, prior commands and an event queue. The method and system is performed by monitoring events and states of the computer system and receiving a processed command corresponding to the spoken command. The processed command is statistically analyzed in light of the system events and states as well as according to an acoustic model. The system then identifies a recognized command corresponding to the spoken command.

公开(公告)号：ES2248018T3

公开(公告)日：2006-03-16

申请号：ES00304415

申请日：2000-05-24

Applicant: IBM

Inventor： BALLARD BARBARA ELAINE ,  LEWIS JAMES R ,  NASSIFF AMADO ,  ORTEGA KERRY A ,  VANBUSKIRK RONALD E ,  WANG HUIFANG

IPC: G10L15/18 ,  G10L15/22 ,  G10L15/26

Abstract: Un método para uso en un programa informático para reconocimiento de voz que funciona en diversos estados y ejecuta un programa para poner en práctica diversos sucesos, para reconocer una instrucción oral, que comprende los pasos de: monitorizar (54) al menos uno de dichos sucesos y estados; recibir una instrucción procesada correspondiente a dicha instrucción oral; analizar (44) dicha instrucción procesada de acuerdo con al menos un modelo acústico a fin de identificar una probable coincidencia acústica; analizar (48) dicha instrucción procesada a fin de identificar una probable coincidencia de contexto usando un modelo estadístico para analizar al menos uno de dichos sucesos y estados de acuerdo con un conjunto finito de instrucciones ponderadas de acuerdo con la probabilidad estadística de sus sucesos correspondientes que ocurren en el estado dado; y proporcionar una instrucción reconocida basada en dichas probables coincidencias acústicas y de contexto.