Method For Speech Recognition From a Partitioned Vocabulary

• US 20080126090A1
• Filed: 10/04/2005
• Published: 05/29/2008
• Est. Priority Date: 11/16/2004
• Status: Active Grant

• Current Assignee: Siemens AG
• Original Assignee: Siemens AG
• Inventors: Kunstmann, Niels
• Granted Patent: US 8,306,820 B2
• Time in Patent Office: Days
• US Class Current: 704/245
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/08 Speech classification or search
• CPC Class Codes: G10L 2015/085 Methods for reducing search complexity, pruning

Abstract

A is recognized using a predefinable vocabulary that is partitioned in sections of phonetically similar words. In a recognition process, first oral input is associated with one of the sections, then the oral input is determined from the vocabulary of the associated section.

Claims

1-11. -11. (canceled)

12. A method for recognizing a spoken input using a predefinable vocabulary, comprising:

storing, on a storage medium, the predefinable vocabulary subdivided into sections of phonetically similar words;

assigning the spoken input to one of the sections; and

determining the spoken input using the phonetically similar words in the one of the sections that has been assigned to the spoken input.

13. The method as claimed in claim 12, wherein a word from the predefinable vocabulary is assigned to at least one section of phonetically similar words.

14. The method as claimed in claim 13,

further comprising characterizing the sections of the phonetically similar words by a representative entry, and

wherein said assigning assigns the spoken input to the one of the sections for which the representative entry is most similar to the spoken input.

15. The method as claimed in claim 14, wherein the spoken input is at least one spoken word.

16. The method as claimed in 14, wherein the spoken input is a series of spoken letters.

17. The method as claimed in claim 16, wherein said storing comprises:

determining distance measures for phonetic similarity of pairs of words; and

subdividing the predefinable vocabulary into the sections of phonetically similar words based on the distance measures using a vector quantization algorithm.

18. The method as claimed in claim 17, wherein said determining of the distance measure for phonetic similarity between the pairs of words comprises:

ranking recognized words with associated ranking value in relation to at least one test utterance, respectively; and

averaging differences of the associated ranking value of the pairs of words for respective test utterances to determine the distance measure between the pairs of words.

19. The method as claimed in claim 17, wherein, said determining of the distance measure for phonetic similarity between the pairs of words comprises:

determining distance values for a similarity of two letter sequences; and

adding the distance values for the distance measure of the two letter sequences.

20. The method as claimed in claim 19, wherein said storing stores the sections of the predefinable vocabulary by different lengths of letter sequences.

21. The method as claimed in claim 20, wherein a Levenshtein distance is used as the distance measure.

22. A computer readable medium encoding a computer program which when executed by a processor causes the processor to perform a method comprising:

storing, on a storage medium, the predefinable vocabulary subdivided into sections of phonetically similar words;

assigning the spoken input to one of the sections; and

determining the spoken input using the phonetically similar words in the one of the sections that has been assigned to the spoken input.