Method of enrolling phone-based speaker specific commands

Details Method of enrolling phone-based speaker specific commands Method of enrolling phone-based speaker specific commands

2000-02-10

2002-04-23

Dorvil, Richemond (Department: 2641)

Data processing: speech signal processing, linguistics, language

Speech signal processing

Recognition

C704S255000

Reexamination Certificate

active

06377924

ABSTRACT:

FIELD OF THE INVENTION
This invention relates to speech recognition and more particularly to enrollment of voice commands which can be recognized to trigger actions.
BACKGROUND OF THE INVENTION
There is a growing demand for voice commands recognition. It has been used for voice name dialing for telephone and user-specific commands such as car controls, computer operations and almost everything that would use the hands to trigger an action. It is even being considered for browsing the Internet. It is the accuracy of the recognition that is important and that is dependent on models generated during enrollment. The recognition of voice commands requires the construction of HMM models on enrollment, during which utterance is recorded and need to build the HMM of the command. Depending on the model level, two types of HMMs can be used. A first and most common type is word-based models which models the whole command (may be several words as a single unit). The second type is phone-based which uses a concatenation of phone-like sub-word units to model a command. The sub-word unit can be represented using speaker-independent HMM as described by N. Jain, R. Cole and E. Barnard in article entitled “Creating Speaker-Specific Phonetic Templates with Speaker-Independent Phonetic Recognizer”; In Proc. IEEE International Conference on Acoustics, Speech, and Signal Processing, page 881-884, Atlanta, May 1996, or speaker specific HMM. While word-based HMMs is easier to train, phone-based HMM has many advantages including various degree of distribution tying and rejection based on phone durations.
SUMMARY OF THE INVENTION
In accordance with one embodiment of the present inventions applicants teach the construction of phone-based HMM for speaker-specific command enrollment comprising the steps of providing a set (H) of speaker-independent phone-based HMMs, providing a gammer (G) comprising a loop of phones with optional between phone silence (BWS) and two utterance (U
1
and U
2
) of the command produced by the enrollment speaker and wherein the first frames of the first utterance containing only background noise, generating a sequence of phone-like unit HMMS and generating the number of HMMs in that sequence.


REFERENCES:
patent: 5317673 (1994-05-01), Cohen et al.
patent: 5572624 (1996-11-01), Sejnoha
patent: 5794192 (1998-08-01), Zhao
patent: 5839105 (1998-11-01), Ostendorf et al.
patent: 5895447 (1999-04-01), Ittycheriah et al.
patent: 5930753 (1999-07-01), Potamianos et al.
patent: 6151573 (2000-11-01), Gong
Neena Jian, et al., “Creating Speaker-Specific Phonetic Templates with a Speaker-Independent Phonetic Recognizer: Implications for Voice Dialing” IEEE, pp. 881-884, 1996.

Affiliated with

Also associated with

Rating

Method of enrolling phone-based speaker specific commands does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Method of enrolling phone-based speaker specific commands, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method of enrolling phone-based speaker specific commands will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-2896928