US6230129B1
Expired / Lapsed
US
⏱️ 0.0 years remaining
Segment-based similarity method for low complexity speech recognizer
Assignee: Panasonic Holdings Corp
•
Filing: 1998-11-25
•
Granted: 2001-05-08
•
Expires: 2018-11-25
Abstract
Abstract A digital word prototype is constructed using one or more speech utterance for a given spoken word or phrase. First, a phone model is used to derive phoneme similarity time series for each of a plurality of phonemes which represent the degree of similarity between the speech utterance and a set of standard phonemes contained in the phone model. Next, the phoneme similarity data is normalized in relation to a non-speech part of the input speech signal. The normalized phoneme similarity data is divided into segments, such that the sum of all normalized phoneme similarity values in a segment are equal for each segment. Next, a word model is constructed from the phoneme similarity data. To do so, within each segment, a summation value is determined by summing over speech frames each of the normalized phoneme similarity values associated with a particular phoneme. In this way, the word model is represented by a vector of summation values that compactly correlate to the normalized phoneme similarity data. Lastly, the results of the individually processed utterances for a given spoken word (i.e., the individual word models) are combined to produce a digital word prototype that electronically represents the given spoken word.
Inventors
- 👤 Philippe R. Morin
- 👤 Ted H. Applebaum
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-11-09 22:46
CPC Classifications
G; G10; G10L; G10L15/00; G10L15/08; G10L15/10; G10L15/02; G10L2015/025
Claims Analysis
Total Claims: 14
Claim 1
Independent Claim
210 words
An apparatus for representing a given speech utterance as a digital word prototype, comprising: a phone model database for storing phone model speech data corresponding to a plurality of phonemes; a phoneme similarity module connected to said phone model database and receptive of said given speech utterance for producing phoneme similarity data indicative of the correlation between the given speech utterance and said phone model speech data with respect to time; a speech normalization module receptive of said phoneme similarity data for normalizing said phoneme similarity data with respect to a non-speech part of the given speech utterance wherein said speech normalization module normalizes said phoneme similarity data by generating an average phoneme similarity value based on the phoneme similarity data in the non-speech part of the given speech utterance and subtracting said average phoneme similarity value from each of the phoneme similarity values associated with said phoneme similarity data; and a word modeling module receptive of said phoneme similarity data and normalized phoneme similarity data for building word model data by segmenting the normalized phoneme similarity data into at least one segment and determining a sum for the normalized phoneme similarity data in each of said segments, thereby electronically representing the given speech utterance using said word model data.
Claim 2
Independent Claim
210 words
An apparatus for representing a given speech utterance as a digital word prototype, comprising: a phone model database for storing phone model speech data corresponding to a plurality of phonemes; a phoneme similarity module connected to said phone model database and receptive of said given speech utterance for producing phoneme similarity data indicative of the correlation between the given speech utterance and said phone model speech data with respect to time; a speech normalization module receptive of said phoneme similarity data for normalizing said phoneme similarity data with respect to a non-speech part of the given speech utterance wherein said speech normalization module normalizes said phoneme similarity data by generating an average phoneme similarity value based on the phoneme similarity data in the non-speech part of the given speech utterance and subtracting said average phoneme similarity value from each of the phoneme similarity values associated with said phoneme similarity data; and a word modeling module receptive of said phoneme similarity data and normalized phoneme similarity data for building word model data by segmenting the normalized phoneme similarity data into at least one segment and determining a sum for the normalized phoneme similarity data in each of said segments, thereby electronically representing the given speech utterance using said word model data.
Claim 3
Dependent Claim
43 words
The apparatus of claim 1 wherein said phoneme similarity data is segmented into at least three segments by said word modeling module, such that the sum of said phoneme similarity data including said plurality of phonemes is equal in each of said segments.
Claim 4
Dependent Claim
31 words
The apparatus of claim 2 wherein said sum being further defined as a Root Sum Square (RSS) value of the phoneme similarity data for each phoneme in each of said segments.
Claim 5
Independent Claim
174 words
An apparatus for representing a given speech utterance as a digital word prototype, comprising: a phone model database for storing phone model speech data corresponding to a plurality of phonemes; a phoneme similarity module connected to said phone model database and receptive of the given speech utterance for producing phoneme similarity data indicative of the correlation between the given speech utterance and said phone model speech data with respect to time; a speech normalization module receptive of said phoneme similarity data for normalizing said phoneme similarity data with respect to a non-speech part of the given speech utterance by averaging the phoneme similarity data in the non-speech part of the given speech utterance signal and subtracting an average phoneme similarity value from each of the phoneme similarity values associated with said phoneme similarity data; and a word modeling module receptive of said phoneme similarity data and said normalized phoneme similarity data for building word model data corresponding to said phoneme similarity data, thereby electronically representing the given speech utterance using said word model data.
Claim 6
Independent Claim
174 words
An apparatus for representing a given speech utterance as a digital word prototype, comprising: a phone model database for storing phone model speech data corresponding to a plurality of phonemes; a phoneme similarity module connected to said phone model database and receptive of the given speech utterance for producing phoneme similarity data indicative of the correlation between the given speech utterance and said phone model speech data with respect to time; a speech normalization module receptive of said phoneme similarity data for normalizing said phoneme similarity data with respect to a non-speech part of the given speech utterance by averaging the phoneme similarity data in the non-speech part of the given speech utterance signal and subtracting an average phoneme similarity value from each of the phoneme similarity values associated with said phoneme similarity data; and a word modeling module receptive of said phoneme similarity data and said normalized phoneme similarity data for building word model data corresponding to said phoneme similarity data, thereby electronically representing the given speech utterance using said word model data.
Claim 7
Dependent Claim
40 words
The apparatus of claim 4 wherein said word modeling module forms word model data by segmenting the normalized phoneme similarity data into at least one segment and determining a sum for the phoneme similarity data in each of said segments.
Claim 8
Dependent Claim
40 words
The apparatus of claim 5 wherein said normalized phoneme similarity data being segmented into at least three segments by said word modeling module, such that the sum of said normalized phoneme similarity data being equal in each of said segments.
Claim 9
Dependent Claim
28 words
The apparatus of claim 5 wherein said sum being further defined as a Root Sum Square (RSS) value for the phoneme similarity data in each of said segments.
Claim 10
Independent Claim
159 words
A method of representing a given spoken word as a digital word prototype, comprising the steps of: providing phone templates representing a database of standard speech; comparing a first training instance with said phone templates to produce first phoneme similarity data as a function of time, said first training instance corresponding to the given spoken word; normalizing said first phoneme similarity data with respect to a non-speech part of said first training instance by averaging the phoneme similarity data in the non-speech part of said first training instance and subtracting an average phoneme similarity value from each phoneme similarity value associated with said first phoneme similarity data; and building first word model data corresponding to said first phoneme similarity data by segmenting said first phoneme similarity data into at least one segment and determining a sum for said first phoneme similarity data in each of said segments, thereby forming the digital word prototype that represents the given spoken word.
Claim 11
Independent Claim
159 words
A method of representing a given spoken word as a digital word prototype, comprising the steps of: providing phone templates representing a database of standard speech; comparing a first training instance with said phone templates to produce first phoneme similarity data as a function of time, said first training instance corresponding to the given spoken word; normalizing said first phoneme similarity data with respect to a non-speech part of said first training instance by averaging the phoneme similarity data in the non-speech part of said first training instance and subtracting an average phoneme similarity value from each phoneme similarity value associated with said first phoneme similarity data; and building first word model data corresponding to said first phoneme similarity data by segmenting said first phoneme similarity data into at least one segment and determining a sum for said first phoneme similarity data in each of said segments, thereby forming the digital word prototype that represents the given spoken word.
Claim 12
Dependent Claim
37 words
The method of claim 8 further comprises segmenting said normalized first phoneme similarity data into at least three segments, such that the sum of said normalized first phoneme similarity data being equal in each of said segments.
Claim 13
Dependent Claim
33 words
The method of claim 8 wherein the step of building word models data further comprises determining a root sum square (RSS) value for said first phoneme similarity data in each of said segments.
Claim 14
Dependent Claim
106 words
The method of claim 8 further comprises: comparing a second training instance with said phone templates to produce second phoneme similarity data as a function of time, said second training instance corresponding to the given spoken word; building second word model data corresponding to said second phoneme similarity data by segmenting said second phoneme similarity data into a plurality of segments and determining a sum for said second phoneme similarity data in each of said segments; and computing a mean sum and a weight for each of said phonemes in each of said segments, thereby forming the digital word prototype that represents the given spoken word.
References Cited (Backward Citations)
US5345536A
Cited
US5513297A
Cited
US5664061A
Cited
US5522011A
Cited
US5806034A
Cited
US5684925A
Cited
US5822728A
Cited
US5825977A
Cited
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 2004-05-07 | 2004-11-06 | 2005-05-08 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2008-05-07 | 2008-11-06 | 2009-05-08 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2012-05-07 | 2012-11-06 | 2013-05-08 | $7700.00 | Window Closed / Lapsed |