US4947436A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Speaker verification using memory address
Assignee: British Telecommunications PLC
•
Filing: 1988-10-07
•
Granted: 1990-08-07
•
Expires: 2008-10-07
Abstract
Abstract Digital representations of samples of frequency spectrum (via filters F1-F16) are used to address a store S. In a training sequence, code words are stored in the store locations indicating whether the associated spectra are common to a number of valid users or are particular to one. For identification, the codewords addressed and read out of the store during the speech of an unknown speaker are counted in counters Z1-Zn and the results used to determine or verify the speaker's identity.
Inventors
- 👤 Alan J. Greaves
- 👤 Paul C. Millar
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-11-04 23:40
CPC Classifications
G; G10; G10L; G10L17/00; G10L17/04
Claims Analysis
Total Claims: 23
Claim 1
Independent Claim
124 words
An apparatus for speaker identification comprising: means for deriving from each of successive simple periods of speech input thereto a digital word representing the frequency spectrum thereof; a store for storing codewords indicating for each digital word the relative probability of the frequency spectrum corresponding to that word occurring in the speech of known speakers which the apparatus has been trained to identify, the digital word output of the deriving means being connected to the address inputs of the store; and control means responsive, during input of speech by a speaker to be identified, to codewords read out from the store to output, in accordance with a predetermined criterion, data indicating which, if any, of the known speakers is deemed to have been identified.
Claim 2
Independent Claim
124 words
An apparatus for speaker identification comprising: means for deriving from each of successive simple periods of speech input thereto a digital word representing the frequency spectrum thereof; a store for storing codewords indicating for each digital word the relative probability of the frequency spectrum corresponding to that word occurring in the speech of known speakers which the apparatus has been trained to identify, the digital word output of the deriving means being connected to the address inputs of the store; and control means responsive, during input of speech by a speaker to be identified, to codewords read out from the store to output, in accordance with a predetermined criterion, data indicating which, if any, of the known speakers is deemed to have been identified.
Claim 3
Dependent Claim
90 words
An apparatus according to claim 1, arranged in operation during a training sequence, for each sample period, as follows: (a) if the number of codewords already stored at the addressed location in the store does not exceed a predetermined number, store at that location a codeword identifying the speaker from whose speech the digital word representing that address was derived, (b) if the number of codewords already stored at the addressed location in the store exceeds the predetermined number, store at that location a codeword reserved for indicating multiple users.
Claim 4
Dependent Claim
13 words
An apparatus according to claim 2 in which the predetermined number is zero.
Claim 5
Dependent Claim
100 words
An apparatus according to claim 1, in which the control means comprises a respective counter means for each known speaker and is arranged in operation to respond to each codeword read out from the store by incrementing the counter corresponding to the known speaker (if any) indicated by that codeword as having a relatively high probability of being associated with the frequency spectrum corresponding to the address from which the codeword was read, and in which the predetermined criterion is that the speaker who shall be deemed to be recognised shall be the one whose counter has the highest count.
Claim 6
Dependent Claim
39 words
An apparatus according to claim 4 in which the predetermined criterion includes the requirement that identification shall not be deemed to have occurred unless the said highest count exceeds the counts of the other counter by a set margin.
Claim 7
Dependent Claim
31 words
An apparatus according to claim 4 in which predetermined criterion includes the requirement that identification shall not be deemed to have occurred unless the said highest count exceeds a threshold value.
Claim 8
Dependent Claim
56 words
An apparatus according to claim 6 in which the control means contain a further counter, wherein the control means is responsive to the occurence of a codeword indicating a low probability of any known speaker, to increment the further counter, and is responsive to the content of the further counter to adjust the said threshold value.
Claim 9
Dependent Claim
65 words
An apparatus according to claim 6 in which said characteristic feature is the frequency spectrum of the input and in which deriving means comprises: filter means for dividing the frequency spectrum of the speech input into a plurality of frequency band outputs, means for measuring the average amplitude of, and sampling those outputs, and coding means for obtaining the digital word from the resulting samples.
Claim 10
Dependent Claim
34 words
An apparartus according to claim 8 in which the coding means is arranged for each band to compare each sample with a threshold value and produce a `1` or `0` signal in dependence thereon.
Claim 11
Dependent Claim
38 words
An apparatus according to claim 8 in which the coding means is arranged for each band to compare each sample with the previous sample obtained for that band and produce a `1` or `0` signal in dependence thereon.
Claim 12
Dependent Claim
37 words
An apparatus according to claim 8 in which the coding means is arranged for each band to compare each sample with the sample obtained for another band and produce a `1` or `0` signal in dependence thereon.
Claim 13
Independent Claim
143 words
An apparatus for receiving spoken input and for recognizing or verifying the identity of the speaker comprising: means for deriving from each of a succession of sample periods of speech input a digital representing the frequency spectrum of the input, means for storing a pluurality of codewords, each codeword indicating for each said digital word the relative probability of the frequency spectrum of said digital word occurring in the speech of known speakers, means for addressing said means for storing codewords, means for connecting the digital word output from said means for deriving to said means for addressing, and control means responsive to codewords accessed by said means for addressing to control the production of data in accordance with a predetermined criterion, said data indicating which, if any, speaker is identified, said control means being responsive when the apparatus is receiving spoken input.
Claim 14
Independent Claim
143 words
An apparatus for receiving spoken input and for recognizing or verifying the identity of the speaker comprising: means for deriving from each of a succession of sample periods of speech input a digital representing the frequency spectrum of the input, means for storing a pluurality of codewords, each codeword indicating for each said digital word the relative probability of the frequency spectrum of said digital word occurring in the speech of known speakers, means for addressing said means for storing codewords, means for connecting the digital word output from said means for deriving to said means for addressing, and control means responsive to codewords accessed by said means for addressing to control the production of data in accordance with a predetermined criterion, said data indicating which, if any, speaker is identified, said control means being responsive when the apparatus is receiving spoken input.
Claim 15
Independent Claim
126 words
A method for identifying the speaker of an unknown input speech signal as one of a group of previously identified speakers, said method comprising: storing in a random access memory stored codewords uniquely corresponding to said previously identified speakers at address locations corresponing to the input speech's frequency spectrum, previously supplied by such speakers, a given one of said codewords being supplied at a memory output in response to input address signals; applying to said memory a succession of addresses corresponding to the frequency spectrum of the unknown input speech signal; and identifying the speaker of the unknown input speech signal as a function of the resulting succession of codewords output from the memory if said speaker is contained in the group of previously identified speakers.
Claim 16
Independent Claim
126 words
A method for identifying the speaker of an unknown input speech signal as one of a group of previously identified speakers, said method comprising: storing in a random access memory stored codewords uniquely corresponding to said previously identified speakers at address locations corresponing to the input speech's frequency spectrum, previously supplied by such speakers, a given one of said codewords being supplied at a memory output in response to input address signals; applying to said memory a succession of addresses corresponding to the frequency spectrum of the unknown input speech signal; and identifying the speaker of the unknown input speech signal as a function of the resulting succession of codewords output from the memory if said speaker is contained in the group of previously identified speakers.
Claim 17
Independent Claim
114 words
A method for receiving spoken input and for recognizing the identity of the speaker comprising the steps of: deriving from each of a succession of input speech samples a digital word representing the frequency spectrum thereof; storing a plurality of codewords in a memory device, each codeword indicating for each said digital word the relative probability of the frequency spectrum of said digital word occurring in the speech of known speakers; connecting the derived digital words to the addressing circuits of said memory device; obtaining a succession of codewords accessed by said addressing circuits; and using said succession of codewords, in accordance with a predetermined criterion, to indicate which, if any, speaker is identified.
Claim 18
Independent Claim
114 words
A method for receiving spoken input and for recognizing the identity of the speaker comprising the steps of: deriving from each of a succession of input speech samples a digital word representing the frequency spectrum thereof; storing a plurality of codewords in a memory device, each codeword indicating for each said digital word the relative probability of the frequency spectrum of said digital word occurring in the speech of known speakers; connecting the derived digital words to the addressing circuits of said memory device; obtaining a succession of codewords accessed by said addressing circuits; and using said succession of codewords, in accordance with a predetermined criterion, to indicate which, if any, speaker is identified.
Claim 19
Independent Claim
219 words
An apparatus for speaker identification comprising: means for deriving from each of successive sample periods of speech input thereto a digital word representing a characteristic feature of the input; a store for storing codewords indicating for each digital word the relative probability of the charateristic feature corresponding to that word occurring in the speech of known speakers which the apparatus has been trained to identify, the digital word output of the deriving means being connected to the address inputs of the store; control means responsive, during input of speech by a speaker to be identified, to codewords read out from the store to output, in accordance with a predetermined criterion, data indicating which, if any, of the known speakers is deemed to have been identified; and wherein said control means includes a respective counter means for each known speaker and is arranged in operation to respond to each codeword read out from the store by incrementing the counter corresponding to the known speaker (if any) indicated by that codeword as having a relatively high probability of being associated with the characteristic feature corresponding to the address from which the codeword was read, and in which the predetermined criterion is that the speaker who shall be deemed to be recognised shall be the one whose counter has the highest count.
Claim 20
Independent Claim
219 words
An apparatus for speaker identification comprising: means for deriving from each of successive sample periods of speech input thereto a digital word representing a characteristic feature of the input; a store for storing codewords indicating for each digital word the relative probability of the charateristic feature corresponding to that word occurring in the speech of known speakers which the apparatus has been trained to identify, the digital word output of the deriving means being connected to the address inputs of the store; control means responsive, during input of speech by a speaker to be identified, to codewords read out from the store to output, in accordance with a predetermined criterion, data indicating which, if any, of the known speakers is deemed to have been identified; and wherein said control means includes a respective counter means for each known speaker and is arranged in operation to respond to each codeword read out from the store by incrementing the counter corresponding to the known speaker (if any) indicated by that codeword as having a relatively high probability of being associated with the characteristic feature corresponding to the address from which the codeword was read, and in which the predetermined criterion is that the speaker who shall be deemed to be recognised shall be the one whose counter has the highest count.
Claim 21
Dependent Claim
39 words
An apparatus according to claim 15 in which the predetermined criterion includes the requirement that identification shall not be deemed to have occurred unless the said highest count exceeds the counts of the other counters by a set margin.
Claim 22
Dependent Claim
31 words
An apparatus according to claim 15 in which predetermined criterion includes the requirement that identification shall not be deemed to have occurred unless the said highest count exceeds a threshold value.
Claim 23
Dependent Claim
52 words
An apparatus according to claim 17 in which the control means contains a further counter, is responsive to the occurrence of a codeword indicating a low probability of any known speaker to increment the further counter, and is responsive to the content of the further counter to adjust the said threshold value.
References Cited (Backward Citations)
US3737580A
Cited
US4100370A
Cited
US4344031A
Cited
US4394538A
Cited
US4720863A
Cited
US4763278A
Cited
US4752958A
Cited
US4783805A
Cited
US4776017A
Cited
US4797927A
Cited
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 1993-08-06 | 1994-02-05 | 1994-08-07 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 1997-08-06 | 1998-02-05 | 1998-08-07 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2001-08-06 | 2002-02-05 | 2002-08-07 | $7700.00 | Window Closed / Lapsed |