Purpose: using the potential function network algorithm of Gauss and a kind of method of learning vector quantization algorithm, it is arranged to generate the best code book for being used for speech recognition for speech recognition. Construction: using the potential function network algorithm of Gauss and a kind of method of learning vector quantization algorithm, one fyord is included by the 6th step for speech recognition. At the first step (S10), by using SPC (linear predictor coefficient) algorithm, a digital data of the characteristic part of digital data is extracted. At second step (S20), by using GPFN (the potential function network of Gauss) algorithm, an output vector x (t) is generated by it. At third step (S30), output vector x (t) inputs LVQ (learning vector quantization) algorithm, and a LVQ algorithm researches are performed. At forward step (S40), a best code book is generated. Recognized in the voice of one the 5th step (S50), the user of a characteristic feature. In the 6th step (S60), recognized by the sound that user expresses.
展开▼