US3770892A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Connected word recognition system
Assignee: International Business Machines Corp
•
Filing: 1972-05-26
•
Granted: 1973-11-06
•
Expires: 1992-05-26
Abstract
Abstract A system is disclosed which recognizes connected or separate spoken words based on the concatenation of steady state sounds produced by a speaker enunciating a given word for which a definitive array of steady state sounds has previously been entered into the system during a learning period.
Inventors
- 👤 G Clapper
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-10-23 14:03
CPC Classifications
G; G10; G10L; G10L15/00
Claims Analysis
Total Claims: 12
Claim 1
Independent Claim
231 words
A method of automatically recognizing spoken words, comprising the steps of: separating full bandwidth electronically manifested and amplified speech signals in an analyzer for passing individual sub-bandwidth components according to frequencies; sensing continuously at a delayed time following a start signal the steady-state output condition signals from said analyzer sub-bands to determine which of said signals are above a continuously defined and varying voltage threshold; storing in a temporary storage device at times determined by clocking signals generated by a clock whose clocking rate is dependent on the speakers''s production of steady-state vocal sounds and on time delays built into said clock which are activated in response to said vocal sounds the patterns of information signals indicative of which of said sensed outputs are above said threshold and also indicative of which of said outputs are below said threshold; comparing said temporarily stored signal patterns with other patterns of signals previously stored in a memory means and identifying the best individual match therebetween for each said temporarily stored pattern; signalling the information results of said comparison step for each said temporarily stored signal pattern; storing sequentially said information signals from said comparison step as uniphone codes for the steady-state speech signals sensed in said sensing step; and recognizing groups of said sequentially stored uniphone codes as words by means of a uniphone sequence-to-word conversion device library, thereby identifying said spoken words.
Claim 2
Independent Claim
231 words
A method of automatically recognizing spoken words, comprising the steps of: separating full bandwidth electronically manifested and amplified speech signals in an analyzer for passing individual sub-bandwidth components according to frequencies; sensing continuously at a delayed time following a start signal the steady-state output condition signals from said analyzer sub-bands to determine which of said signals are above a continuously defined and varying voltage threshold; storing in a temporary storage device at times determined by clocking signals generated by a clock whose clocking rate is dependent on the speakers''s production of steady-state vocal sounds and on time delays built into said clock which are activated in response to said vocal sounds the patterns of information signals indicative of which of said sensed outputs are above said threshold and also indicative of which of said outputs are below said threshold; comparing said temporarily stored signal patterns with other patterns of signals previously stored in a memory means and identifying the best individual match therebetween for each said temporarily stored pattern; signalling the information results of said comparison step for each said temporarily stored signal pattern; storing sequentially said information signals from said comparison step as uniphone codes for the steady-state speech signals sensed in said sensing step; and recognizing groups of said sequentially stored uniphone codes as words by means of a uniphone sequence-to-word conversion device library, thereby identifying said spoken words.
Claim 3
Dependent Claim
33 words
A method as described in claim 1, further including the step of: encoding said recognized words from said converting step into coded form for transmission out of the system as recognized word codes.
Claim 4
Dependent Claim
41 words
A method as described in claim 1, wherein: said separating, storing and comparing steps are coordinated and controlled by clocking signals generated by a clock at times derived in response to the integrated vocal production of speech signals by the speaker.
Claim 5
Dependent Claim
38 words
A method as defined in claim 3, further comprising a step of: stopping said clock and said operations controlled thereby whenever an absence of signals is detected and by restarting said clock upon the resUmption of input signals.
Claim 6
Dependent Claim
40 words
A method of claim 3, further comprising a step of: changing said clocking signals to a slower rate whenever fricative sounds of a duration longer than 50 milliseconds are detected so as to reduce redundant samples of the same sound.
Claim 7
Dependent Claim
338 words
A word recognition system, comprising: transducer means for electrically manifesting voice signals for recognition; frequency analysis means connected to said transducer means for separating said voice signals into a plurality of frequency band components; amplification means in association with said frequency analysis means for amplifying said frequency band components; selection and signalling means connected to the output of said amplification means for selecting from among said amplified frequency band components those bands whose band electrical energy content exceeds a threshold level which varies for each frequency band in proportion to the amount of energy being passed in adjacent, sub-adjacent and any further removed adjacent frequency bands and for signalling which of said bands are so selected thus forming a band selection signal pattern; synchronization and control means for coordinating the operations of the system by generating controlling clocking signals, said means being connected to said frequency analysis and selection means for the receipt of signals thereform and responsive thereto for generating said clocking signals to control the operation of the following system elements, comprising; first storage means connected to said selection means for temporarily storing said selection signal pattern outputs therefrom; second storage means for storing a plurality of signal patterns expected from the output of said selection means; comparison, decision and signalling means connected to said first and second storage means for comparing band selection signal patterns from said selection means with said patterns stored in said second storage means and for deciding which comparison results in the closest match and for signalling the identity of the pattern in said second storage means so chosen; third storage means connected to the output of said comparison means for temporarily storing the identities of a plurality of said chosen patterns for input, under the control of said synchronization and control means, to the following elements; conversion means connected to said third storage means for converting pluralities of pattern identities thereform into word identities as recognized words upon the receipt of a clocking signal from said synchronization and control means.
Claim 8
Dependent Claim
338 words
A word recognition system, comprising: transducer means for electrically manifesting voice signals for recognition; frequency analysis means connected to said transducer means for separating said voice signals into a plurality of frequency band components; amplification means in association with said frequency analysis means for amplifying said frequency band components; selection and signalling means connected to the output of said amplification means for selecting from among said amplified frequency band components those bands whose band electrical energy content exceeds a threshold level which varies for each frequency band in proportion to the amount of energy being passed in adjacent, sub-adjacent and any further removed adjacent frequency bands and for signalling which of said bands are so selected thus forming a band selection signal pattern; synchronization and control means for coordinating the operations of the system by generating controlling clocking signals, said means being connected to said frequency analysis and selection means for the receipt of signals thereform and responsive thereto for generating said clocking signals to control the operation of the following system elements, comprising; first storage means connected to said selection means for temporarily storing said selection signal pattern outputs therefrom; second storage means for storing a plurality of signal patterns expected from the output of said selection means; comparison, decision and signalling means connected to said first and second storage means for comparing band selection signal patterns from said selection means with said patterns stored in said second storage means and for deciding which comparison results in the closest match and for signalling the identity of the pattern in said second storage means so chosen; third storage means connected to the output of said comparison means for temporarily storing the identities of a plurality of said chosen patterns for input, under the control of said synchronization and control means, to the following elements; conversion means connected to said third storage means for converting pluralities of pattern identities thereform into word identities as recognized words upon the receipt of a clocking signal from said synchronization and control means.
Claim 9
Dependent Claim
84 words
A word recognition system as described in claim 6, further comprising: word detection and encoding means connected to said conversion means for the receipt of word identities therefrom and for encoding the same; and a gated output storage means connected to said synchronization and control means and to said word detection and encoding means for the receipt of encoded words therefrom and for storing the same until said synchronization control means gates the output from said output storage means as an encoded recognized word.
Claim 10
Dependent Claim
132 words
A word recognition system as described in claim 7, wherein: said frequency analysis means comprises a series of contiguous sub-bandpass filters whose combined bandpass encompasses the range of human voice signals; said amplification means for amplifying said frequency band components comprises a logarithmic amplifier connected to the input of each said sub-bandpass filter and logarithmic amplifiers connected to the outputs of said filters whose sub-bandpass frequencies lie below 4K Hz; and said selection and signalling means comprises a voltage threshold comparator connected to the amplified output of each said sub-bandpass segment of said frequency analysis means, said comparator having a resistive network on its input to connect it with its adjacent, sub-adjacent and any further removed comparators and to prOportionately raise the threshold voltage level for each said comparator so connected therewith.
Claim 11
Dependent Claim
44 words
A word recognition system as described in claim 8, wherein: said comparison, decision and signalling means is an adaptive electronic memory comprising a plurality of electronic templates and associated decision circuits for signalling which of said templates contains the pattern having the best match.
Claim 12
Dependent Claim
40 words
A word recognition system as described in claim 9, wherein: said conversion means is a plugboard to which pluralities of identified uniphone patterns are separately wired to form the words which are desired for outputs in response to spoken words.
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) | 1976-11-05 | 1977-05-07 | 1977-11-06 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 1980-11-05 | 1981-05-07 | 1981-11-06 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 1984-11-05 | 1985-05-07 | 1985-11-06 | $7700.00 | Window Closed / Lapsed |