Voice recognition apparatus and method of controlling the same

1. A method of controlling a voice recognition apparatus, the method comprising:

generating a plurality of recognition tokens based on a voice signal received by the voice recognition apparatus;

searching for a similar token and a peripheral token for each of the recognition tokens, from among a plurality of pre-stored tokens, the similar token being a pre-stored token having a high degree of similarity with the recognition token and the peripheral token being a pre-stored token previously used with the recognition token;

generating a token network using the recognition tokens, the similar tokens, and the peripheral tokens;

calculating a transition probability between the tokens constituting the token network; and

generating text data for the received voice signal using the calculated transition probability,

wherein, in response to a first recognition token not being present among the plurality of pre-stored tokens, the searching includes searching for the peripheral token corresponding to a second recognition token located just after the first recognition token.

2. The method as claimed in claim 1 , wherein the plurality of pre-stored tokens are generated by dividing at least one of a content list, a corpus, and an utterance log by predetermined units of tokens.

3. The method as claimed in claim 1 , wherein the searching includes searching for the similar token using a string searching algorithm in response to the recognition token not being presented among the plurality of pre-stored tokens.

4. The method as claimed in claim 1 , which the generating of the plurality of recognition tokens includes:

generating intermediate text data for the received voice signal by recognizing the received voice signal; and

generating the plurality of recognition tokens by dividing the generated intermediate text data for the received voice signal in predetermined units,

wherein the generating of the token network includes arranging the similar tokens and the peripheral tokens to correspond to the plurality of recognition tokens.

5. The method as claimed in claim 1 , further comprising:

calculating a probability in which the text data occurs; and

outputting a request message to repeat the voice signal in response to the probability in which the text data occurs being less than or equal to a predetermined value.

6. The method as claimed in claim 2 , wherein the generating of the text data for the received voice signal includes:

calculating the transition probability between the tokens by assigning predetermined weights to language models; and

searching for an optimum path on the token network based on the calculated transition probability between the tokens.

7. The method as claimed in claim 6 , wherein the searching for the optimum path includes using at least one of a Viterbi algorithm and an a-star algorithm.

8. A voice recognition apparatus comprising:

a voice recognizer configured to receive a voice signal;

a storage unit configured to store a plurality of tokens;

a token network generator configured to generate a plurality of recognition tokens based on the received voice signal, to search for a similar token and a peripheral token for each of the recognition tokens, from among the plurality of pre-stored tokens, the similar token being a pre-stored token having a high degree of similarity with the recognition token and the peripheral token being a pre-stored token previously used with the recognition token, and to generate a token network using the recognition tokens, the similar tokens, and the peripheral tokens; and

a processor configured to control the token network generator to generate the token network in response to the voice signal being received by the voice recognizer, to calculate a transition probability between the tokens constituting the token network, and to generate text data for the received voice signal using the calculated transition probability,

wherein, in response to a first recognition token not being present among the plurality of pre-stored tokens, the token network generator searches for the peripheral token corresponding to a second recognition token located just after the first recognition.

9. The voice recognition apparatus as claimed in claim 8 , wherein the storage unit stores the plurality of tokens, generated by dividing at least one of a content list, a corpus, and an utterance log by predetermined units of tokens, and a degree of similarity between the tokens, and

the token network generator searches for the similar tokens using the degree of similarity between the tokens in response to the recognition token being presented among the plurality of stored tokens.

10. The voice recognition apparatus as claimed in claim 8 , wherein the token network generator searches for the similar tokens using a string searching algorithm in response to the recognition token not being presented among the plurality of stored tokens.

11. The voice recognition apparatus as claimed in claim 8 , wherein the processor generates intermediate text data for the received voice signal in response to the voice signal being received by the voice recognizer, and

the token network generator generates the plurality of recognition tokens by dividing the generated intermediate text data for the received voice signal in predetermined units, and generates the token network by arranging the similar tokens and the peripheral tokens to correspond to the plurality of recognition tokens.

12. The voice recognition apparatus as claimed in claim 11 , wherein the processor calculates the transition probability between the tokens by assigning predetermined weights to language models, and generates the text data for the received voice signal by searching for an optimum path on the token network based on the calculated transition probability between the tokens.

13. The voice recognition apparatus as claimed in claim 12 , wherein the processor searches for the optimum path using at least one of a Viterbi algorithm and an a-star algorithm.

14. The voice recognition apparatus as claimed in claim 8 , further comprising at least one of a display configured to display an image and an audio output unit configured to output sound,

wherein the processor calculates a probability in which the text data occurs, and controls the at least one of the display and the audio output unit to output a request message to repeat the voice signal in response to the probability in which the text data occurs being less than or equal to a predetermined value.

15. A non-transitory computer-readable medium including a program code configured to execute the method of claim 1 .

16. A method of voice recognition, comprising:

generating, using a processor, words of text based on a voice signal received by a voice recognition apparatus, as an initial recognition;

adding the words of the initial recognition to a word network;

for each word of the initial recognition, adding a similar-sounding word to the word network, and selectively adding a peripheral word previously used in conjunction with the word of the initial recognition to the word network if the peripheral word exists;

calculating a transition probability between each word in the to the word network; and

generating a final recognition of the received voice signal using the calculated transition probability.