Reference pattern adapting device trainable by a small number of

Details Reference pattern adapting device trainable by a small number of Reference pattern adapting device trainable by a small number of

1991-07-31

1993-12-28

Knepper, David D.

G10L 506

Patent

active

052747379

ABSTRACT:
In a reference pattern adapting device for adapting reference pattern parameter groups (1-M) to adapted reference pattern parameter groups (1-M) by using training patterns (1-L) derived from a training pattern signal (TP). A similarity calculating unit (15) calculates a group of primary degrees of similarity between (a) each of M reference pattern element groups representative of reference patterns (output of a vector group memory 11) and (b) each of M reference pattern parameter groups (output of pattern parameter memory 12). A difference calculating unit (16) calculates a group of differences between reference pattern elements of each of the M reference pattern element groups and training pattern elements of each of L training pattern element groups representative of the training patterns (1-L). A first calculating unit (21) calculates a primary difference value by using the group of primary degrees of similarity and the group of differences. A second calculating unit (22) calculates a group of primary mean values by using each of L reference pattern parameter groups and the group of differences. The resulting group of primary mean values is output to an identifier (25) as one of L adapted pattern parameter groups.

REFERENCES:
L. R. Rabiner et al., "An Introduction to Hidden Markov Models", IEEE ASSP Magazine, Jan. 1986, pp. 4-16.
S. Furui, "A Training Procedure . . . Recognition Systems", IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. ASSP-28, No. 2, Apr. 1980, pp. 129-136.
Shinoda et al., "Speaker Adaptation for Demi-Syllable Based Continuous Density HMM", IEEE Int'l Conf. on Acoustics, Speech and Signal Processing, vol. 2, May 1991, pp. 857-860.
Furui, "Unsupervised Speaker Adaptation Method Based on Hierarchical Spectral Clustering", Int'l Conf. on Acoustics, Speech and Signal Processing, vol. 1, May 22, 1989, pp. 286-689.
Bonneau, "Vector Quantization for Speaker Adaptation", Int'l Conf. on Acoustics, Speech and Signal Processing, vol. 3, Apr. 6, 1987, pp. 1434-1437.
Shiraki, "Speaker Adaptation Algorithms Based on Piece-Wise Moving Adaptive Segment", Int'l Conf. on Acoustics, Speech and Signal Processing, vol. 2, Apr. 3, 1990, pp. 657-660.

Affiliated with

Also associated with

Rating

Reference pattern adapting device trainable by a small number of does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Reference pattern adapting device trainable by a small number of, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Reference pattern adapting device trainable by a small number of will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-1549894