System and method for multi-lingual speech recognition
A system for multi-lingual speech recognition. The inventive system includes a speech modeling engine, a speech search engine, and a decision reaction engine. The speech modeling engine receives and transfers a mixed multi-lingual speech signal into speech features. The speech search engine locates and compares candidate data sets. The decision reaction engine selects resulting speech models from the candidate speech models and generates a speech command.
1. A system for multi-lingual speech recognition, comprising:
a digital signal processing unit;
a speech modeling system, receiving and transferring a mixed multi-lingual speech signal into a plurality of speech features;
a multi-lingual baseform mapping engine, comparing a plurality of multi-lingual query commands to obtain a plurality of multi-lingual baseforms;
a cross-lingual diphone model generation engine executed by the digital signal processing unit, coupled to the multi-lingual baseform mapping engine, selecting and combining the multi-lingual baseforms, further comprising:
fixing left contexts of the multi-lingual baseforms and mapping right contexts of the multi-lingual baseforms to obtain a mapping result;
fixing right context and mapping the left contexts of the multi-lingual baseforms to obtain the mapping result if the contexts of the multi-lingual baseforms mapping fails; and
obtaining the multi-lingual context-speech mapping data according to the mapping result;
storing the multi-lingual context-speech mapping data in a multi-lingual model database;
a speech search engine, coupled to the speech modeling engine, receiving the speech features, and locating and comparing a plurality of candidate data sets corresponding to the speech features according to the multi-lingual model database to find match probability of a plurality of candidate speech models of the candidate data sets; and
a decision reaction engine, coupled to the speech search engine, selecting a plurality of resulting speech models corresponding to the speech features according to the match probability from the candidate speech models to generate a speech command.
2. The system as claimed in claim 1 , wherein the multi-lingual model database comprises a plurality of multi-lingual anti-models.
3. The system as claimed in claim 2 , further comprising:
at least one uni-lingual anti-model generation engine, receiving a plurality of multi-lingual query commands to generate a plurality of uni-lingual anti-models corresponding to specific languages; and
an anti-model combination engine, coupled to the uni-lingual anti-model generation engine, calculating the uni-lingual anti-models to generate the multi-lingual anti-models.
4. The system as claimed in claim 1 , wherein the speech search engine locates and compares the candidate data sets, further referring the connecting sequences of the speech features and a speech rule database.
5. A method for multi-lingual speech recognition, comprising the steps of:
performing the following steps by a digital signal processing system;
transferring a mixed multi-lingual speech signal into a plurality of speech features;
comparing a plurality of multi-lingual query commands to obtain a plurality of multi-lingual baseforms;
selecting and combining the multi-lingual baseforms, comprising:
fixing left contexts of the multi-lingual baseforms and mapping right contexts of the multi-lingual baseforms to obtain a mapping result;
fixing right context and mapping the left contexts of the multi-lingual baseforms to obtain the mapping result if the right contexts of the multi-lingual baseforms mapping fails; and
obtaining the multi-lingual context-speech mapping data according to the mapping result;
storing the multi-lingual context-speech mapping data in a multi-lingual model database;
locating and comparing a plurality of candidate data sets corresponding to the speech features according to the multi-lingual model database to find match probability of a plurality of candidate speech models of the candidate data sets; and
selecting a plurality of resulting speech models corresponding to the speech features from the candidate speech models according to the match probability to generate a speech command.
6. The method as claimed in claim 5 , wherein the multi-lingual model database comprises a plurality of multi-lingual anti-models.
7. The method as claimed in claim 6 , further comprising the steps of:
receiving a plurality of multi-lingual query commands corresponding to specific languages and generate a plurality of uni-lingual anti-models; and
combining the uni-lingual anti-models to generate the multi-lingual anti-model.
8. The method as claimed in claim 5 , wherein locating and comparison of the candidate data sets further refers the connecting sequences of the speech features and a speech rule database.