Method, medium, and apparatus recognizing speech
AssigneeSamsung Electronics Co., Ltd.
InventorsIcksang Han, Sangbae Jeong, Jeongsu Kim
Life of this patent
Estimating the term…
Abstract
An apparatus, medium, and method recognizing speech. The method may include the calculating of scores indicating the degree of similarity between a characteristic of an input speech and characteristics of speech models based on the degree of similarity between the length of each phoneme included in an input speech and the length of phonemes included in each speech model, and determining a speech model with the highest score among the scores to be the corresponding recognized speech for the input speech. By doing so, the speech recognition rate may be greatly enhanced and when an input speech includes continuous identical phonemes the word error rate (WER) may be greatly reduced.
Claims24
A speech recognition method, comprising:
- calculating scores indicating a degree of similarity between a characteristic of an input speech and characteristics of speech models, of a plurality of speech models, based on a degree of similarity between lengths of phonemes included in the input speech and lengths of phonemes included in the speech models;
- and
- indicating a speech model having a high score among the calculated scores as being a speech model matching the input speech.
A speech recognition apparatus, comprising:
- a score calculation unit to calculate scores indicating a degree of similarity between a characteristic of an input speech and characteristics of speech models, of a plurality of speech models, based on a degree of similarity between lengths of phonemes included in the input speech and lengths of phonemes included in the speech models;
- and
- a recognized speech calculation unit to identify a speech model having a high score among the calculated scores as a speech model matching the input speech.
The speech recognition apparatus of claim 9, wherein the high score is a highest score among the calculated scores.
A medium comprising computer readable code to implement a speech recognition method, wherein the speech recognition method comprises:
- calculating scores indicating a degree of similarity between a characteristic of an input speech and characteristics of speech models, of a plurality of speech models, based on a degree of similarity between lengths of phonemes included in the input speech and lengths of phonemes included in the speech models;
- and
- indicating a speech model, of the plurality of speech models, having a high score among the calculated scores as being a speech model matching the input speech.
A method for calculating a score, comprising:
- calculating correlation coefficients indicating a degree of similarity between lengths of phonemes included in an input speech and lengths of phonemes included in candidate speeches, of a plurality of candidate speeches;
- and
- calculating and outputting scores indicating a degree of similarity of a characteristic of the input speech and characteristics of the candidate speeches based on the calculated correlation coefficients.
An apparatus for calculating a score comprising:
- a correlation coefficient calculation unit to calculate correlation coefficients indicating a degree of similarity between lengths of phonemes included in an input speech and lengths of phonemes included in candidate speeches, of a plurality of speeches;
- and
- a score calculation unit to calculate scores indicating a degree of similarity of a characteristic of the input speech and characteristics of the candidate speeches based on the calculated correlation coefficients.
A medium comprising computer readable code to implement a score calculation method, wherein the score calculation method comprises:
- calculating correlation coefficients indicating a degree of similarity between lengths of phonemes included in an input speech and lengths of phonemes included in candidate speeches, of a plurality of candidate speeches;
- and
- calculating and outputting scores indicating a degree of similarity of a characteristic of the input speech and characteristics of the candidate speeches based on the calculated correlation coefficients.
Dependencies are inferred from the claim text.