Grammar-based checksum constraints for high performance speech recognition circuit

• US 5,222,187 A
• Filed: 12/29/1989
• Issued: 06/22/1993
• Est. Priority Date: 12/29/1989
• Status: Expired due to Term

• Current Assignee: Texas Instruments, Inc.
• Original Assignee: Texas Instruments, Inc.
• Inventors: Hemphill, Charles T., Doddington, George R.
• Primary Examiner(s): Fleming, Michael R.
• Assistant Examiner(s): Doerrler, Michelle
• Application Number:
  US07/458,860
• Time in Patent Office: 1,271 Days
• Field of Search: 381/41-43, 381/51, 364/513.5, 371/65, 371/37, 371/39, 371/40, 395/2
• US Class Current: 704/200
• CPC Class Codes: G10L 15/193 Formal grammars, e.g. finite state automata, context free grammars or word networks
• CPC Class Codes: G10L 17/22 Interactive procedures; Man-machine interfaces

Abstract

A speech system recognizes words from a spoken phrase that conform to checksum constraints. Grammar rules are applied to hypothesize words according to the checksum constraints. The checksum associated with the phrase is thus inherent in the grammar. Sentences which do not meet a predetermined checksum constraint are not valid under the grammar rules and are therefore inherently rejected. The checksum constraints result in increased recognition accuracy.

Claims

1. A method of recognizing speech comprising the steps of:

receiving spoken input;

observing words from said spoken input;

conforming a sequence of said observed words to one or more checksum constraints; and

recognizing said conformed sequence by controlling such speech recognition with a grammar embedding said checksum constraints.

2. The method of claim 1 wherein said recognizing step includes the steps of:

hypothesizing words from said spoken input to form a hypothesized phrase; and

applying a grammar rule to said hypothesized words, said grammar rule operable to implement a checksum constraint on the hypothesized phrase.

3. The method of claim 2 wherein each hypothesized word comprises a number.

4. The method of claim 2 wherein said step of applying includes the step of applying a grammar rule associated with a first state and with said hypothesized words such that a second state is determined, said second state indicative of a checksum value.

5. The method of claim 2 and further comprising the step of computing a probability for each hypothesized word to aid implementation of said checksum constraint on said hypothesized phrase.

6. The method of claim 1 wherein said controlling step comprises the step of controlling the speech recognition using a finite-state automaton embedding said checksum constraints.

7. The method of claim 1 wherein said controlling step comprises the step of controlling the speech recognition using a unification grammar embedding said checksum constraints.

8. Circuitry for recognizing speech comprising:

circuitry for receiving spoken input;

circuitry for observing words from said spoken input;

circuitry for conforming a sequence of said observed words to one or more checksum constraints; and

circuitry for recognizing said conformed sequence by controlling such speech recognition with a grammar embedding said checksum constraints.

9. The circuitry of claim 8 wherein said recognizing circuitry comprises:

circuitry for hypothesizing words from said spoken input to form a hypothesized phrase; and

circuitry for applying a grammar rule to said hypothesized words operable to implement a checksum constraint on the hypothesized phrase.

10. The circuitry of claim 9 wherein said circuitry for hypothesizing words comprises circuitry for hypothesizing numbers.

11. The circuitry of claim 9 wherein said applying circuitry comprises circuitry for applying a grammar rule associated with a first state and with an observed word such that a second state is determined, said second state indicative of a checksum value.

12. The circuitry of claim 9 and further comprising circuitry for computing a probability for each hypothesized work to aid implementation of said checksum constraint on said hypothesized phrase.

13. The circuitry of claim 8 wherein said controlling circuitry comprisess circuitry for controlling the speech recognition using a finite-state grammar embedding said checksum constraints.

14. The circuitry of claim 8 wherein said controlling circuitry comprises circuitry for controlling the speech recognition using a unification grammar embedding said constraints.

15. A method of recognizing speech from a spoken phrase comprising a plurality of words in a grammar-based system, comprising the steps of:

receiving said spoken phrase;

activating a first state;

hypothesizing a word from the spoken phrase;

applying a grammar rule associated with said first state to said hypothesized word and an observed word such that a second state is determined, said second state indicative of a checksum value; and

recognizing speech by controlling such speech recognition with a grammar embedding checksum constraints which employ said checksum value.

16. The method of claim 15 wherein each hypothesized word comprises a number.

17. The method of claim 15 and further comprising the steps of:

repeating said steps of hypothesizing and applying until the end of the spoken phrase is reached, wherein said steps of hypothesizing words and applying associated grammar rules thereto are performed in order to determine states associated with a next word position and respective of said hypothesized words.

18. The method of claim 17 and further comprising the step of computing a probability for each hypothesized phrase to aid in said step of recognizing speech.

19. The method of claim 18 and further comprising activating each of said determined successive states after applying the grammar rule.

20. The method of claim 19 and further comprising the step of comparing the probabilities of hypothetical phrases when a determined state has previously been activated in conjunction with another hypothesized phrase.

21. The method of claim 15 wherein said step of hypothesizing further comprises hypothesizing phrases of successive said hypothesized words from said spoken phrase;

and

said step of applying further comprises applying grammar rules to each of said successive hypothesized words to determine successive states associated with successive word positions, each successive state indicative of a checksum value.

22. The method of claim 21 wherein each hypothesized phrase comprises a plurality of numbers.

23. The method of claim 15 and further comprising the step of computing a probability for each hypothesized word to aid in said step of recognizing speech.

24. The method of claim 15 wherein said grammar rules comprise finite-state automaton transitions.

25. The method of claim 15 wherein said grammar rules comprise unification grammar rules.

26. Circuitry for recognizing speech from a spoken phrase comprising a plurality of words, comprising:

circuitry for hypothesizing words from the spoken phrase;

circuitry for applying grammar rules to determine a transition from a source state to a destination state responsive to a hypothesized word, such that a checksum may be determined from said destination state; and

circuitry for recognizing speech by controlling such speech recognition with a grammar embedding checksum constraints by employing said checksum.

27. The circuitry of claim 26 wherein each hypothesized work comprises a number.

28. The circuitry of claim 26 wherein each state is associated with a work position within a hypothesized phrase comprised of successive hypothesized words and a checksum value for the hypothesized phrase.

29. The circuitry of claim 28 wherein said hypothesizing circuitry is operable to hypothesize a plurality of hypothesized phrases for each spoken phrase.

30. The circuitry of claim 29 and further comprising circuitry for computing a probability for each hypothesized work to aid said circuitry for recognizing speech.

31. The circuitry of claim 30 and further comprising circuitry for computing a probability for each hypothesized phrase to aid said circuitry for recognizing speech.

32. The circuitry of claim 29 and further comprising circuitry for computing a probabilities of hypothetical phrases having a common destination state to aid said circuitry for recognizing.