VOICE RECOGNITION APPARATUS, VOICE RECOGNITION METHOD AND RECORDING MEDIUM

• US 20100010813A1
• Filed: 04/30/2009
• Published: 01/14/2010
• Est. Priority Date: 07/10/2008
• Status: Active Grant

• Current Assignee: Fujitsu Limited
• Original Assignee: Fujitsu Limited
• Inventors: HARADA, Shouji
• Granted Patent: US 8,271,282 B2
• Time in Patent Office: Days
• US Class Current: 704/249
• CPC Class Codes: G10L 15/08 Speech classification or search
• CPC Class Codes: G10L 2015/025 Phonemes, fenemes or fenones being the recognition units

Abstract

A voice recognition apparatus includes an extraction unit extracting a feature amount from a voice signal, a word dictionary storing a plurality of recognition words; a reject word generation unit storing reject words in the word dictionary in association with the recognition words and a collation unit calculating a degree of similarity between the voice signal and each of the recognition words and reject words stored in the word dictionary by using the feature amount extracted by the extraction unit, determining whether or not a word having a high calculated degree of similarity corresponds to a reject word, when the word is determined as the reject word, excluding the recognition word stored in the word dictionary in association with the reject word from a result of recognition, and outputting a recognition word having a high calculated degree of similarity as a result of recognition.

Claims

1. A voice recognition apparatus comprising:

an extraction unit extracting a feature amount from a voice signal;

a word dictionary storing a plurality of recognition words;

a reject word generation unit storing reject words in the word dictionary in association with the recognition words; and

a collation unit calculating a degree of similarity between the voice signal and each of the recognition words and reject words stored in the word dictionary by using the feature amount extracted by the extraction unit, determining whether or not a word having a high calculated degree of similarity corresponds to a reject word, when the word is determined as the reject word, excluding the recognition word stored in the word dictionary in association with the reject word from a result of recognition, and outputting a recognition word having a high calculated degree of similarity as a result of recognition.

2. The voice recognition apparatus according to claim 1, further comprising:

an input unit inputting a plurality of voice signals with respect to a plurality of reject word candidates relating to a recognition word to which a reject word is added, to the extraction unit;

wherein the extraction unit is designed to extract a feature amount from each of the inputted voice signals;

wherein

the collation unit calculates the degree of similarity between the voice signal and each of the recognition words stored in the word dictionary by using the feature amount extracted by the extraction unit, and is designed to output the result of recognition based upon the calculated degree of similarity; and

the reject word generation unit calculates the number of voice signals each corresponding to a recognition word to which the reject word is added as the result of recognition for each of the reject word candidates based upon the result of recognition outputted by the collation unit, and stores a reject word candidate having a large number of calculations carried out by the reject word generation unit as a reject word relating to the recognition word to which the reject word is added.

3. The voice recognition apparatus according to claim 2, wherein the reject word generation unit is designed to store the reject word candidate having a number of calculations carried out by the reject word generation unit that exceeds a given number in the word dictionary as a reject word relating to the recognition word to which the reject word is added.

4. The voice recognition apparatus according to claim 2, wherein the reject word generation unit is designed to store one portion of the reject word candidate in the word dictionary as a reject word relating to the recognition word to which the reject word is added.

5. The voice recognition apparatus according to claim 3, wherein the reject word generation unit is designed to store one portion of the reject word candidate in the word dictionary as a reject word relating to the recognition word to which the reject word is added.

6. The voice recognition apparatus according to claim 1, further comprising:

an input unit inputting a plurality of voice signals with respect to a plurality of reject word candidates relating to a recognition word to which a reject word is added, to the extraction unit;

wherein the extraction unit is designed to extract a feature amount from each of the inputted voice signals, wherein

the collation unit calculates the degree of similarity between the voice signal and each of the recognition words stored in the word dictionary by using the feature amount extracted by the extraction unit, and is designed to output the result of recognition based upon the calculated degree of similarity; and

the reject word generation unit generates an acoustic model from a voice signal having a high degree of similarity to the recognition word to which the reject word is added, calculated by the collation unit, and stores the generated acoustic model in the word dictionary as a reject word relating to the recognition word.

7. A voice recognition apparatus comprising:

an extraction unit extracting a feature amount from a voice signal, and storing the feature amount in a storage unit;

a word dictionary storing a plurality of recognition words;

a reject word generation unit storing reject words in the word dictionary in association with the recognition words; and

a collation unit calculating a degree of similarity between the voice signal and each of the recognition words and reject words stored in the word dictionary by using the feature amount stored in the storage unit by the extraction unit, determining whether or not a word having a high calculated degree of similarity corresponds to a reject word, when the word is determined as the reject word, excluding the recognition word stored in the word dictionary in association with the reject word from the result of recognition, and outputting a recognition word having a high calculated degree of similarity as a result of recognition.

8. A voice recognition method allowing a voice recognition apparatus including a word dictionary that stores a plurality of recognition words and reject words to output a recognition result based on a feature amount extracted from a voice signal, said voice recognition method comprising:

calculating a degree of similarity between the voice signal and each of recognition words and reject words stored in the word dictionary by using the extracted feature amount;

determining whether or not a word having a high calculated degree of similarity corresponds to a reject word;

when the word is determined as the reject word, excluding the recognition word stored in the word dictionary in association with the reject word from a result of recognition.

9. A computer-readable recording medium storing a computer program comprising:

causing a computer to extract a feature amount from a voice signal;

causing the computer to calculate a degree of similarity between the voice signal and each of recognition words and reject words stored in a word dictionary by using the extracted feature amount;

causing the computer to determine whether or not a word having a high calculated degree of similarity corresponds to a reject word;

when the word is determined as the reject word, causing the computer to exclude the recognition word stored in the word dictionary in association with the reject word from a result of recognition.