Miyuki Tanaka

US Patent:

61732583, Jan 9, 2001

Filed:

Oct 22, 1998

Appl. No.:

9/177461

Inventors:

Xavier Menendez-Pidal - Los Gatos CA
Miyuki Tanaka - Campbell CA
Ruxin Chen - San Jose CA
Duanpei Wu - Sunnyvale CA

Assignee:

Sony Corporation - Tokyo
Sony Electronics Inc. - Park Ridge NJ

International Classification:

G10L 506
G10L 900
G10L 302

US Classification:

704233

Abstract:

A method for reducing noise distortions in a speech recognition system comprises a feature extractor that includes a noise-suppressor, one or more time cosine transforms, and a normalizer. The noise-suppressor preferably performs a spectral subtraction process early in the feature extraction procedure. The time cosine transforms preferably operate in a centered-mode to each perform a transformation in the time domain. The normalizer calculates and utilizes normalization values to generate normalized features for speech recognition. The calculated normalization values preferably include mean values, left variances and right variances.

US Patent:

62161032, Apr 10, 2001

Filed:

Oct 20, 1997

Appl. No.:

8/957875

Inventors:

Duanpei Wu - Sunnyvale CA
Miyuki Tanaka - Campbell CA
Ruxin Chen - San Jose CA
Lex Olorenshaw - Madera CA

Assignee:

Sony Corporation - Tokyo
Sony Electronics Inc. - Park Ridge NJ

International Classification:

G01L 300

US Classification:

704253

Abstract:

A method for implementing a speech recognition system for use during conditions with background noise includes the steps of calculating, in real-time, sequential short-term delta energy parameters for speech energy from a spoken utterance, determining threshold values in the speech energy, and identifying a beginning point and an ending point for the spoken utterance based on the relationship between the threshold values and the short-term delta energy parameters.

US Patent:

62301225, May 8, 2001

Filed:

Oct 21, 1998

Appl. No.:

9/176178

Inventors:

Duanpei Wu - Sunnyvale CA
Miyuki Tanaka - Campbell CA
Mariscela Amador-Hernandez - San Jose CA

Assignee:

Sony Corporation - Tokyo
Sony Electronics Inc. - Park Ridge NJ

International Classification:

G10L 2102

US Classification:

704226

Abstract:

A method for effectively suppressing background noise in a speech detection system comprises a filter bank for separating source speech data into discrete frequency sub-bands to generate filtered channel energy, and a noise suppressor for weighting the frequency sub-bands to improve the signal-to-noise ratio of the resultant noise-suppressed channel energy. The noise suppressor preferably includes a subspace module for using a Karhunen-Loeve transformation to create a subspace based on the background noise, a projection module for generating projected channel energy by projecting the filtered channel energy onto the created subspace, and a weighting module for applying calculated weighting values to the projected channel energy to generate the noise-suppressed channel energy.

US Patent:

60061863, Dec 21, 1999

Filed:

Oct 16, 1997

Appl. No.:

8/953026

Inventors:

Ruxin Chen - San Jose CA
Miyuki Tanaka - Campbell CA
Duanpei Wu - Sunnyvale CA
Lex S. Olorenshaw - Corte Madera CA

Assignee:

Sony Corporation - Tokyo
Sony Electronics, Inc. - Park Ridge NJ

International Classification:

G10L 708

US Classification:

704254

Abstract:

A method and an apparatus for a parameter sharing speech recognition system are provided. Speech signals are received into a processor of a speech recognition system. The speech signals are processed using a speech recognition system hosting a shared hidden Markov model (HMM) produced by generating a number of phoneme models, some of which are shared. The phoneme models are generated by retaining as a separate phoneme model any triphone model having a number of trained frames available that exceeds a prespecified threshold. A shared phoneme model is generated to represent each of the groups of triphone phoneme models for which the number of trained frames having a common biphone exceed the prespecified threshold. A shared phoneme model is generated to represent each of the groups of triphone phoneme models for which the number of trained frames having an equivalent effect on a phonemic context exceed the prespecified threshold. A shared phoneme model is generated to represent each of the groups of triphone phoneme models having the same center context.