US5937381A
Expired / Lapsed
US
⏱️ 0.0 years remaining
System for voice verification of telephone transactions
Assignee: ITT Defense and Electronic Inc
•
Filing: 1996-04-10
•
Granted: 1999-08-10
•
Expires: 2016-04-10
Abstract
Abstract A system and a method is disclosed for verifying a voice of a user conducting a telephone transaction. The system and method includes a mechanism for prompting the user to speak in a limited vocabulary. A feature extractor converts the limited vocabulary into a plurality of speech frames. A pre-processor is coupled to the feature extractor for processing the plurality of speech frames to produce a plurality of processed frames. The processing includes frame selection, which eliminates each of the plurality of speech frames having an absence of words. A Viterbi decoder is also coupled to said feature extractor for assigning a frame label to each of the plurality of speech frames to produce a plurality of frame labels. The processed frames and frame labels are then combined to produce a voice model, which includes each of the plurality of frame labels that correspond to the number of plurality of processed frames. A mechanism is also provided for comparing the voice model with the claimant's voice model, derived during a previous enrollment session. The voice model also is compared with an alternate voice model set, derived during previous enrollment sessions. The identity claimed is accepted if the voice model matches the claimant's voice model better than the alternative voice model set.
Inventors
- 👤 William Yee-Ming Huang
- 👤 Lawrence George Bahler
- 👤 Alan Lawrence Higgins
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-11-02 07:08
CPC Classifications
G; G10; G10L; G10L17/00; G10L17/22; G10L17/24
Claims Analysis
Total Claims: 23
Claim 1
Independent Claim
179 words
A system for verifying a voice of a user conducting a telephone transaction, comprising: a means for prompting said user to speak in a limited vocabulary; a feature extractor coupled to said limited vocabulary for converting said limited vocabulary into a plurality of speech frames; a pre-processor coupled to said feature extractor for processing said plurality of speech frames to produce a plurality of processed frames, said processing including frame selection means for selecting those frames having power levels falling within a nominal range of acceptable energies, which eliminates each of said plurality of speech frames which are below or exceed said nominal range, including those frames having an absence of words; A Viterbi decoder coupled to said feature extractor for assigning a frame label to each of said plurality of speech frames producing a plurality of frame labels; and a means for combining said plurality of processed frames with said plurality of frame labels to produce a voice model including each of the plurality of frame labels that correspond to the number of said plurality of processed frames.
Claim 2
Independent Claim
179 words
A system for verifying a voice of a user conducting a telephone transaction, comprising: a means for prompting said user to speak in a limited vocabulary; a feature extractor coupled to said limited vocabulary for converting said limited vocabulary into a plurality of speech frames; a pre-processor coupled to said feature extractor for processing said plurality of speech frames to produce a plurality of processed frames, said processing including frame selection means for selecting those frames having power levels falling within a nominal range of acceptable energies, which eliminates each of said plurality of speech frames which are below or exceed said nominal range, including those frames having an absence of words; A Viterbi decoder coupled to said feature extractor for assigning a frame label to each of said plurality of speech frames producing a plurality of frame labels; and a means for combining said plurality of processed frames with said plurality of frame labels to produce a voice model including each of the plurality of frame labels that correspond to the number of said plurality of processed frames.
Claim 3
Dependent Claim
35 words
The system of claim 1, wherein said limited vocabulary includes at least one enrollment phrase selected from the group of 24 combinations of the words 4,6,7 & 9 arranged in a double two-digit number combination.
Claim 4
Dependent Claim
17 words
The system of claim 1, wherein said model is a HMM model including a plurality of nodes.
Claim 5
Dependent Claim
23 words
The system of claim 1, wherein said processing further includes channel equalization of said plurality of speech frames by a blind deconvolution method.
Claim 6
Dependent Claim
34 words
The system of claim 1, wherein said frame labels assigned to each of said plurality of speech frames are generated by comparing said plurality of speech frames to a model of said limited vocabulary.
Claim 7
Dependent Claim
18 words
The system of claim 5, wherein each said frame label identifies the most probable node of said HMM.
Claim 8
Dependent Claim
29 words
The system of claim 1, wherein said voice model is produced by pairing each of said plurality of frame labels to a sub-set of said plurality of processed frames.
Claim 9
Dependent Claim
113 words
The system of claim 1, which further includes: means for measuring a nearest neighbor distance between said voice model and a claimant voice model, said nearest neighbor distances calculated by comparing the individual frames of said voice model and said claimant voice model that have a corresponding frame label; and means for comparing said nearest neighbor distance with a predetermined threshold, wherein said comparing means is operative to produce a signal selected from the group consisting of an accept signal and a reject signal, wherein said reject signal is produced if said nearest neighbor distance exceeds said threshold and said accept signal is produced if said nearest neighbor distance is within said threshold.
Claim 10
Dependent Claim
142 words
The system of claim 1, which further includes: means for measuring a plurality of nearest neighbor distances between said voice model and a plurality of voice models, said nearest neighbor distances calculated by comparing the individual frames of said voice model and said plurality of voice model that have a corresponding label, said plurality of voice models including said voice model and an alternate voice model set; means for approximating a likelihood ratio from said plurality of nearest neighbor distances; and means for comparing said likelihood ratio with a predetermined threshold, wherein said comparing means is operative to produce a signal selected from the group consisting of an accept signal and a reject signal, wherein said reject signal is produced if said likelihood ratio exceeds said threshold and said accept signal is produced if said likelihood ratio is less than said threshold.
Claim 11
Dependent Claim
26 words
The system of claim 9, wherein said alternate voice model set is a single alternate voice model set utilized for a large group of enrolled users.
Claim 12
Independent Claim
144 words
A method for verifying a voice of a user conducting a telephone transaction, comprising: prompting said user to speak in a limited vocabulary; converting said limited vocabulary into a plurality of speech frames; processing said plurality of speech frames to produce a plurality of processed frames, said processing including selecting those frames having power levels falling within a nominal range of acceptable energies which eliminates each of said plurality of speech frames which are below or exceed said nominal range, including those frames having an absence of words; assigning a frame label to each of said plurality of speech frames to produce a plurality of frame labels; and combining said plurality of processed frames with said plurality of frame labels to produce a voice model including each of the plurality of frame labels that correspond to the number of said plurality of processed frames.
Claim 13
Independent Claim
144 words
A method for verifying a voice of a user conducting a telephone transaction, comprising: prompting said user to speak in a limited vocabulary; converting said limited vocabulary into a plurality of speech frames; processing said plurality of speech frames to produce a plurality of processed frames, said processing including selecting those frames having power levels falling within a nominal range of acceptable energies which eliminates each of said plurality of speech frames which are below or exceed said nominal range, including those frames having an absence of words; assigning a frame label to each of said plurality of speech frames to produce a plurality of frame labels; and combining said plurality of processed frames with said plurality of frame labels to produce a voice model including each of the plurality of frame labels that correspond to the number of said plurality of processed frames.
Claim 14
Dependent Claim
38 words
The method of claim 11, wherein said limited vocabulary includes at least one enrollment phrase selected from the group of 24 combinations of the words 4,6,7 & 9 that can be arranged in a double two-digit number combination.
Claim 15
Dependent Claim
17 words
The method of claim 11, wherein said model is a HMM model including a plurality of nodes.
Claim 16
Dependent Claim
23 words
The method of claim 11, wherein said processing further includes channel equalization of said plurality of speech frames by a blind deconvolution method.
Claim 17
Dependent Claim
34 words
The method of claim 11, wherein said frame label assigned to each of said plurality of speech frames is generated by comparing said plurality of speech frames to a model of said limited vocabulary.
Claim 18
Dependent Claim
17 words
The method of claim 15, wherein each said label identifies a most probable node of said HMM.
Claim 19
Dependent Claim
29 words
The method of claim 16, wherein said voice model is produced by pairing each of said plurality of frame labels to a sub-set of said plurality of processed frames.
Claim 20
Dependent Claim
104 words
The method of claim 11, which further includes: measuring a nearest neighbor distance between said voice model and a claimant voice model, said nearest neighbor distance calculated by comparing the individual frames of said voice model and said claimant voice model that have a corresponding label; and comparing said nearest neighbor distance with a predetermined threshold in order to produce a signal selected from the group consisting of an accept signal and a reject signal, wherein said reject signal is produced if said nearest neighbor distance exceeds said threshold and said accept signal is produced if said nearest neighbor distance is within said threshold.
Claim 21
Dependent Claim
131 words
The method of claim 11, which further includes: measuring a plurality of nearest neighbor distances between said voice model and a plurality of voice models, said nearest neighbor distances calculated by comparing the individual frames of said voice model and said plurality of voice model that have a corresponding label, said plurality of voice models including said voice model and an alternate voice model set; producing a likelihood ratio from said plurality of nearest neighbor distances; and comparing said likelihood ratio with a predetermined threshold in order to produce a signal selected from the group consisting of an accept signal and a reject signal, wherein said reject signal is produced if said likelihood ratio exceeds said threshold and said accept signal is produced if said likelihood ratio is within said threshold.
Claim 22
Dependent Claim
26 words
The method of claim 19, wherein said alternate voice model set is a single alternate voice model set utilized for a large group of enrolled users.
Claim 23
Dependent Claim
129 words
The method of claim 20, wherein said alternate voice model set is formed by: producing a number of voice models for a large number of users; entering a first voice model from said number of voice models into said voice model set; for each of the voice models, from said number of voice models, measuring nearest neighbor distances between another voice model from said number of voice models and each voice model in said alternate voice model set; comparing said measured distances between the another voice model and each voice model in said alternate voice model set with a predetermined threshold; and adding said another voice model to said alternate voice model set only if all said measured distances associated with said another voice model exceed said predetermined threshold.
References Cited (Backward Citations)
US5216720A
Cited
US5159638A
Cited
US5297194A
Cited
US5499288A
Cited
US5295223A
Cited
US5167004A
Cited
US5465290A
Cited
US5271088A
Cited
US5390278A
Cited
US5765127A
Cited
US5809455A
Cited
US5339385A
Cited
US5450524A
Cited
US5341456A
Cited
US5459814A
Cited
US5649055A
Cited
US5430827A
Cited
US5414755A
Cited
US5774849A
Cited
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 2002-08-09 | 2003-02-08 | 2003-08-10 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2006-08-09 | 2007-02-08 | 2007-08-10 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2010-08-09 | 2011-02-08 | 2011-08-10 | $7700.00 | Window Closed / Lapsed |