The Key Technology Research Of Voice-Print Recognition

The Voice-Print recognition is one of the biology technology, a technology of automatically identify the identification of speaker according to the voice parameters which can show the Speaker's characteristics of physiology and behaviors reflected by voice wave. The theory of the Voice-Print recognition is to distill the unique voice characteristics of the speaker and saves them into data base through voice stylebook recorded in advance, and matches them with a great deal characteristics in the data base when applied to judge the speaker's identity.The most important technologies are characteristics pick-up and mode matching, and the task of characteristics pick-up is to distill the acoustics and language characteristics of the speaker which have strongly separability and high stability. Through analyzing various parameters which can show characteristics of Voice-Print, it's discovered that LPC coefficient based on full track top model and the Mel Cepstrum coefficient using characteristics of tonality can well incarnate the speaker's characteristics. Simultaneity the article analyzes capability of each window function used when adding window during characteristics pick-up, and considering that it may engender energy leakage when adding window to cut off the signal, the article chooses Hamming Window as the window function.The task of the mode matching is to do the similar matching of the diagnostic mode of training and recognition. The KMP pattern matching algorithm is a classical arithmetic, but there is a defect that there exists repeated comparison between text string and pattern string. The article designs a new arithmetic aiming at this shortage. Not like the KMP which puts the pointer of pattern string backward, this arithmetic puts forward the pointer of text string for k units according to eigenvalue k of the current character in pattern string, and makes the pointer of pattern string point to the beginning when current comparison fails, and then starts a new comparison, so it can reduce the time of comparison and improve the speed of matching. It is proved by experiments that this arithmetic can improve efficiency.