Speech recognition apparatus and method for matching inputted speech and a word generated from stored referenced phoneme data

• US 6,236,964 B1
• Filed: 02/14/1994
• Issued: 05/22/2001
• Est. Priority Date: 02/01/1990
• Status: Expired due to Fees

• Current Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Original Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Inventors: Sakurai, Atsushi, Kosaka, Tetsuo, Tamura, Junichi
• Primary Examiner(s): Dorvil, Richemond
• Application Number:
  US08/194,807
• Time in Patent Office: 2,654 Days
• Field of Search: 395/2--, 381/41-45, 704/231, 704/237, 704/236, 704/238, 704/239, 704/240, 704/241, 704/243, 704/244, 704/246, 704/250, 704/251, 704/255, 704/254, 704/249, 704/256, 704/257
• US Class Current: 704/254
• CPC Class Codes: G10L 15/10 using distance or distortion measures between unknown speech and reference templates
• CPC Class Codes: G10L 2015/025 Phonemes, fenemes or fenones being the recognition units

Abstract

A speech recognition method and apparatus in which a speech section is sliced by the unit of a word by spotting and candidate words are selected. Next, in a second stage, matching is conducted by the unit of a phoneme. Consequently, selection of the candidate words and slicing of the speech section can be performed concurrently. Furthermore, narrowing of the candidate words is facilitated. Furthermore, since reference phoneme patterns under a plurality of environments are prepared, recognition of an input speech under a larger number of conditions is possible using a smaller amount of data when compared with the case in which reference word patterns under a plurality of environments are prepared.

Claims

1. A speech recognition method comprising the steps of:

inputting speech into a speech recognition apparatus;

discriminating a candidate word included in the inputted speech based on a similarity obtained by matching the inputted speech and reference words stored in a word dictionary, and assigning a candidate word code to the candidate word;

decomposing the candidate word discriminated in said discriminating step into a plurality of obtained phonemes in accordance with the candidate word code, and assigning a phoneme code to each of the plurality of phonemes;

generating a word by connecting a plurality of reference phoneme data stored in a phoneme dictionary, each being selected to correspond to each of the phoneme codes assigned in said decomposing step; and

recognizing a word included in the input speech based on a similarity obtained by matching the inputted speech and the generated word.

2. A speech recognition apparatus comprising:

a word dictionary for storing a plurality of reference words;

input means for inputting speech;

discriminating means for discriminating a candidate word based on a similarity obtained by matching the inputted speech and the plurality of reference words stored in said word dictionary, and for assigning a candidate word code to the candidate word;

decomposing means for decomposing the candidate word discriminated by said discriminating means into a plurality of phonemes obtained in accordance with the candidate word code, and for assigning a phoneme code to each of the plurality of phonemes;

a phoneme dictionary for storing reference phoneme data;

means for reading out reference phoneme data from said phoneme dictionary corresponding to each of the plurality of phoneme codes, that are products of the decomposing and assigned by said decomposing means, and for generating a word by connecting the reference phoneme data that is read out; and

recognizing means for recognizing a word included in the inputted speech based on a similarity obtained by matching the inputted speech and the generated word.