# US 5,704,005 A · Speech recognition apparatus and word dictionary therefor > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 5,704,005 A - Document ID: US5704005A - Status: Granted - Kind: A - Application number: 08/377,727 - Filed: 1995-01-25 - Granted: 1997-12-30 - Earliest claimed: 1994-01-28 - Assignees: Fujitsu Limited - Inventors: Hitoshi Iwamida - Web page: https://patentgrep.com/patent/US5704005A ## CPC - G10L 15/187: Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams - G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING ## Abstract A speech recognition apparatus is provided with a word dictionary managing subword sequences of words that have possibility of being input together with information indicating importance of each of subwords in each subword sequence which forms a word, where the importance indicates whether or not the subwords of the word are an important factor when recognizing the word, a managing part for managing standard values of acoustic feature patterns of each of the subwords, a generating part for successively reading the words stored in the word dictionary and for generating acoustic feature patterns of each of read subwords using the standard values managed by the managing part, and a collating part for calculating and evaluating distances between an acoustic feature pattern of an input word that is to be recognized and each of the acoustic feature patterns generated by the generating part by referring to the information. ## Claims ### Claim 1 A speech recognition apparatus comprising: a word dictionary storing and managing subword sequences of words and information indicating an importance of each subword in each word, said importance indicating whether or not each subword of each word is an important factor when recognizing each word, said importance being independent for each of the words; managing means for managing standard values of acoustic feature patterns of each subword; generating means, coupled to said word dictionary and said managing means, for successively reading the subword sequences of the words stored in said word dictionary and for generating acoustic feature patterns of each word using the standard values managed by said managing means; and collating means coupled to said word dictionary and said generating means, for calculating and evaluating distances between an acoustic feature pattern of an input word that is to be recognized and each of the acoustic feature patterns generated by said generating means by referring to said information. ### Claim 2 (depends on 1) The speech recognition apparatus as claimed in claim 1, wherein said word dictionary stores said information indicating the importance of each subword in two or more levels of importance. ### Claim 3 (depends on 2) The speech recognition apparatus as claimed in claim 2, wherein said collating means comprises correction means for correcting said distances based on said levels of importance so that the corrected distances become larger for more important subwords. ### Claim 4 (depends on 2) The speech recognition apparatus as claimed in claim 2, wherein a first level of importance indicates a subword with an accent as being important, and a second level of importance indicates a subword with no accent as not being important. ### Claim 5 (depends on 2) The speech recognition apparatus as claimed in claim 2, wherein a first level of importance indicates a subword with a strongest accent as being most important, a second level of importance indicates a subword with a second strongest accent as being second important, and a third level of importance indicates a subword as being less important than the first and second levels of importance. ### Claim 6 (depends on 2) The speech recognition apparatus as claimed in claim 2, wherein a first level of importance indicates a subword that is unlikely to become unvoiced when pronounced as being important, and a second level of importance indicates a subword that is likely to become unvoiced when pronounced as not being important. ### Claim 7 (depends on 2) The speech recognition apparatus as claimed in claim 2, which further comprises: output means, coupled to said collating means, for outputting a word having a smallest sum total of distances of the acoustic feature patterns of the subwords from the acoustic feature pattern of the input word as a recognition result of the input word. ### Claim 8 (depends on 1) The speech recognition apparatus as claimed in claim 1, which further comprises: output means, coupled to said collating means, for outputting a word having a smallest sum total of distances of the acoustic feature patterns of the subwords from the acoustic feature pattern of the input word as a recognition result of the input word. ### Claim 9 (depends on 1) The speech recognition apparatus as claimed in claim 1, wherein each of said subwords is formed by a syllable. ### Claim 10 (depends on 1) The speech recognition apparatus as claimed in claim 1, wherein each of said subwords is formed by one or a plurality of phonetic symbols. ### Claim 11 A word dictionary adapted for use on a speech apparatus which recognizes an input word by collating an acoustic feature pattern of the input word with acoustic feature patterns of words stored in a word dictionary, said word dictionary comprising: a recording medium; subword sequences of words stored in said recording medium; and information indicating an importance of each subword in each of the subword sequences which form the words, said information being stored in said recording medium together with said subword sequences, said importance indicating whether or not each subword of each word is an important factor when recognizing the input word on the speech recognition apparatus, and said importance being independent for each word. ### Claim 12 (depends on 11) The word dictionary as claimed in claim 11, wherein said information indicating the importance of the subwords is stored in said recording medium in two or more levels of importance. ### Claim 13 (depends on 12) The word dictionary as claimed in claim 12, wherein a first level of importance indicates a subword with an accent as being important, and a second level of importance indicates a subword with no accent as not being important. ### Claim 14 (depends on 12) The word dictionary as claimed in claim 12, wherein a first level of importance indicates a subword with a strongest accent as being most important, a second level of importance indicates a subword with a second strongest accent as being second important, and a third level of importance indicates a subword as being less important than the first and second levels of importance. ### Claim 15 (depends on 12) The word dictionary as claimed in claim 12, wherein a first level of importance indicates a subword that is unlikely to become unvoiced when pronounced as being important, and a second level of importance indicates a subword that is likely to become unvoiced when pronounced as not being important. ### Claim 16 (depends on 11) The word dictionary as claimed in claim 11, wherein each of said subwords is formed by a syllable. ### Claim 17 (depends on 11) The word dictionary as claimed in claim 11, wherein each of said subwords is formed by one or a plurality of phonetic symbols. ### Claim 18 A speech recognition apparatus comprising: a word dictionary storing and managing subword sequences of words, and information indicating an importance of each subword in each subword sequence, the importance indicating whether the subwords of each word are an important factor for word recognition, and the importance including at least an accent of each subword; managing means for managing standard values of acoustic feature patterns of each subword; generating means, coupled to said word dictionary and said managing means, for successively reading the subword sequences stored in said word dictionary and for generating acoustic feature patterns of each word using the standard values managed by said managing means; and collating means coupled to said word dictionary and said generating means, for calculating and evaluating distances between an acoustic feature pattern of an input word that is to be recognized and each of the acoustic feature patterns generated by said generating means by referring to said information. ### Claim 19 A word dictionary adapted for use on a speech recognition apparatus which recognizes an input word by collating an acoustic feature pattern of the input word with acoustic feature patterns of words stored in the word dictionary, said word dictionary comprising: a recording medium; subword sequences of words stored in said recording medium; and information indicating an importance of each subword in each subword sequence stored in said recording medium together with said subword sequences, said importance indicating whether each subword of each word is an important factor for word recognition with the speech recognition apparatus, and said importance including at least an accent of each subword.