US5450523A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Training module for estimating mixture Gaussian densities for speech unit models in speech recognition systems
Assignee: Panasonic Holdings Corp
•
Filing: 1993-06-01
•
Granted: 1995-09-12
•
Expires: 2013-06-01
Abstract
Abstract A model-training module generates mixture Gaussian density models from speech training data for continuous, or isolated word speech recognition systems. Speech feature sequences are labeled into segments of states of speech units using Viterbi-decoding based optimized segmentation algorithm. Each segment is modeled by a Gaussian density, and the parameters are estimated by sample mean and sample covariance. A mixture Gaussian density is generated for each state of each speech unit by merging the Gaussian densities of all the segments with the same corresponding label. The resulting number of mixture components is proportional to the dispersion and sample size of the training data. A single, fully merged, Gaussian density is also generated for each state of each speech unit. The covariance matrices of the mixture components are selectively smoothed by a measure of relative sharpness of the Gaussian density and the smoothing can also be done blockwise. The weights of the mixture components are set uniformly initially, and are reestimated using a segmental-average procedure. The weighting coefficients, together with the Gaussian densities, then become the models of speech units for use in speech recognition.
Inventors
- 👤 Yunxin Zhao
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-11-03 00:02
CPC Classifications
G; G10; G10L; G10L15/00; G10L15/06; G10L15/063; G10L15/02; G10L15/08; G10L15/14; G10L15/142; G10L15/144
Claims Analysis
Total Claims: 13
Claim 1
Independent Claim
229 words
A speech recognition system including a training module for estimating mixture Gaussian densities for use in speech-unit models of the speech recognition system, said speech recognition system comprising: means for receiving an input speech signal; segmentation means for segmenting said received input speech signal into a plurality of speech units, each speech unit having a plurality of states, each state of each speech unit representing a segment of said signal, each segment including a plurality of frames; means for grouping all segments corresponding to like speech unit states for like speech units into respective sets of segments for all states of all speech units; means for estimating parameters defining a Gaussian density for each segment of each set of segments; means for merging the Gaussian densities in each set of segments into a mixture Gaussian density through successive merging of a plurality (k) of pairs of said Gaussian densities of each set of segments into (k) Gaussian densities, wherein k is an integer, wherein a plurality of mixture Gaussian densities, one for each set of segments, are created; means for estimating weights for each one of said mixture Gaussian densities, one weight for each Gaussian density in a mixture Gaussian density; and means for using the mixture Gaussian densities and the estimated weights therefore from said means for estimating weights for identifying speech units of additional input speech signals.
Claim 2
Dependent Claim
229 words
A speech recognition system including a training module for estimating mixture Gaussian densities for use in speech-unit models of the speech recognition system, said speech recognition system comprising: means for receiving an input speech signal; segmentation means for segmenting said received input speech signal into a plurality of speech units, each speech unit having a plurality of states, each state of each speech unit representing a segment of said signal, each segment including a plurality of frames; means for grouping all segments corresponding to like speech unit states for like speech units into respective sets of segments for all states of all speech units; means for estimating parameters defining a Gaussian density for each segment of each set of segments; means for merging the Gaussian densities in each set of segments into a mixture Gaussian density through successive merging of a plurality (k) of pairs of said Gaussian densities of each set of segments into (k) Gaussian densities, wherein k is an integer, wherein a plurality of mixture Gaussian densities, one for each set of segments, are created; means for estimating weights for each one of said mixture Gaussian densities, one weight for each Gaussian density in a mixture Gaussian density; and means for using the mixture Gaussian densities and the estimated weights therefore from said means for estimating weights for identifying speech units of additional input speech signals.
Claim 3
Dependent Claim
72 words
The speech recognition system of claim 1, wherein said means for merging comprises: means for calculating distances between all pairs of Gaussian densities in a set; means for sorting the calculated distances into a list of distances in increasing order to find Gaussian density pairs separated by small distances; and means for merging said Gaussian pairs from the start of the list until k pairs are merged into k respective Gaussian densities.
Claim 4
Dependent Claim
98 words
The speech recognition system of claim 2, wherein said means for merging said Gaussian densities into a mixture Gaussian density includes means for merging a single pair of Gaussian densities at each time said means for merging a single pair of Gaussian densities at each time comprises: means for calculating, for each pair of Gaussian densities in the set, a distance separating the Gaussian densities of the pair; means for comparing the calculated distances to find a Gaussian density pair with a minimum separation distance; and means for merging said minimum distance pair into a single Gaussian density.
Claim 5
Dependent Claim
110 words
The speech recognition system of claim 2, wherein said minimum distance Gaussian densities of the pair are represented by indices i* and j*, and a said merged Gaussian density is parameterized by μ and C, and wherein said means for merging said minimum distance Gaussian pair comprises: means for calculating μ, wherein μ is given by: μ=λ.sub.i.μ.sub.i.+λ.sub.j.μ.sub.j.; and means for calculating C, wherein C is given by: C=λ.sub.i.C.sub.i.+λ.sub.j.C.sub.j.+λ.sub.i.λ.sub.j.(μ.sub.i.-μ.sub.j.) (μi.μ.sub.j.)'; wherein λ i* is given by: ##EQU19## and wherein λ j* is given by: ##EQU20## wherein L i* and L j* represent the number of frames in each subset of segments used to estimate the Gaussian densities i* and j* .
Claim 6
Dependent Claim
49 words
The training module of claim 2, wherein a distance d(i,j) between a pair of Gaussian densities i and j is given by: ##EQU21## wherein L i and L j represent the number of frames in each subset of segments used to estimate the Gaussian densities i and j .
Claim 7
Dependent Claim
89 words
The speech recognition system of claim 1 wherein said parameters defining a Gaussian density for each segment comprise μ and C, and wherein said means for estimating said parameters comprises: means for calculating μ, wherein μ is given by: ##EQU22## wherein i is a segment index, L i is the number of frames in segment i, and x l .sup.(i) is the lth feature vector in segment i; and means for calculating C, wherein C is given by: ##EQU23## where the symbol ' denotes the transpose of a vector.
Claim 8
Independent Claim
199 words
A method for speech recognition, said speech recognition method comprising the steps of: receiving an input speech signal; segmenting the input speech signal into a plurality of speech units, each speech unit having a plurality of states, each state of each speech unit representing a segment of said signal, each segment including a plurality of frames; grouping all segments corresponding to like speech unit states for like speech units into respective sets of segments for all states of all speech units; estimating parameters defining a Gaussian density for each segment of each set of segments; merging the Gaussian densities in each set of segments into a mixture Gaussian density through successive merging of a plurality (k) of pairs of said Gaussian densities of each set of segments into (k) Gaussian densities, wherein k is an integer, wherein a plurality of mixture Gaussian densities, one for each set of segments, are created; estimating the weights for each one of said mixture Gaussian densities, one weight for each Gaussian density in a mixture Gaussian density; and using the mixture Gaussian densities and the estimated weights therefore from said means for estimating weight for identifying speech units of additional input speech signals.
Claim 9
Independent Claim
199 words
A method for speech recognition, said speech recognition method comprising the steps of: receiving an input speech signal; segmenting the input speech signal into a plurality of speech units, each speech unit having a plurality of states, each state of each speech unit representing a segment of said signal, each segment including a plurality of frames; grouping all segments corresponding to like speech unit states for like speech units into respective sets of segments for all states of all speech units; estimating parameters defining a Gaussian density for each segment of each set of segments; merging the Gaussian densities in each set of segments into a mixture Gaussian density through successive merging of a plurality (k) of pairs of said Gaussian densities of each set of segments into (k) Gaussian densities, wherein k is an integer, wherein a plurality of mixture Gaussian densities, one for each set of segments, are created; estimating the weights for each one of said mixture Gaussian densities, one weight for each Gaussian density in a mixture Gaussian density; and using the mixture Gaussian densities and the estimated weights therefore from said means for estimating weight for identifying speech units of additional input speech signals.
Claim 10
Dependent Claim
76 words
The method for speech recognition of claim 7 wherein said step of merging (k) pairs of Gaussian densities comprises the steps of: calculating distances between all pairs of Gaussian densities in a set; sorting the calculated distances into a list of distances in increasing order to find Gaussian density pairs separate by small distances; and merging small distance Gaussian pairs from the start of the list until k pairs are merged into k respective Gaussian densities.
Claim 11
Dependent Claim
122 words
The method for speech recognition of claim 8 wherein said minimum distance Gaussian densities of the pair are represented by indices i* and j*, and a said merged Gaussian density is parameterized by μ and C, and wherein said step of merging said minimum distance Gaussian pair comprises the steps of: calculating μ, wherein μ is given by: μ=λ.sub.i* μ.sub.i* +λ.sub.j* μ.sub.j* ; and calculating C, wherein C is given by: C=λ.sub.i* C.sub.i* +λ.sub.j* C.sub.j* +λ.sub.i* λ.sub.j* (μ.sub.i* -μ.sub.j*) (μ.sub.i* -μ.sub.j*)'; wherein λ i* is given by: ##EQU24## and wherein λ j* is given by: ##EQU25## wherein L i* and L j* represent the number of frames in each subset of segments used to estimate the Gaussian densities i* and j* .
Claim 12
Dependent Claim
98 words
The method of speech recognition of claim 7, wherein said step of merging said Gaussian densities into a mixture Gaussian density includes a step of merging a single pair of Gaussian densities at each time, wherein said step of merging a single pair of Gaussian densities at each time includes the steps of: calculating, for each pair of Gaussian densities in the set, a distance separating the Gaussian densities of the pair; comparing the calculated distances to find a Gaussian density pair with a minimum separation distance; and merging said minimum distance pair into a single Gaussian density.
Claim 13
Dependent Claim
51 words
The method of speech recognition of claim 7, wherein a distance d(i,j) between a pair of Gaussian densities i and j is given by: ##EQU26## wherein L i and L j represent the number of frames in each subset of segments used to estimate the Gaussian densities i and j .
References Cited (Backward Citations)
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 1998-09-11 | 1999-03-13 | 1999-09-12 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2002-09-11 | 2003-03-13 | 2003-09-12 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2006-09-11 | 2007-03-13 | 2007-09-12 | $7700.00 | Window Closed / Lapsed |