Research On The Speaker Recognition Based On VQ And ANN

Starting in 1930's, speaker recognition has increasingly become a hotspot of research since 1960's. It can be applied to a number of fields, such as security, justice, military affairs, finance and services. Because of that, lots of scientific researchers are involved in the research, making great development. However it is not ripe very much.This paper is mainly about a text-dependent speaker recognition system based on vector quantification (VQ) and (ANN)methods, a text-independent speaker recognition system based on Gaussian mixture models. We use LPC-derived cepstral coefficients, pitches as the feature parameter set. Through the test of a speech library composed of 10 speakers and 1800 speeches.Mainly works in this paper : i), The problem of the feature parameter pick-up in speaker recognition, expound the vocal track, LPC analysis, LPC-derived cepstral coefficient, cepstral coefficient, Mel-cepstral coefficient in detail. ii), It mainly Introduce some different methods of the speaker recognition. The usability of the vector quantization (VQ) technique and fuzzy VQ technique in speaker recognition is dissertation in detail. The essence of VQ is to use several special features to represent the whole features in solution so that to achieve the aim of coding and compressing. At the same time, It mainly expounds the base thought and disposal methods of the genetic-algorithm. While if we combine Genetic-algorithm with VQ technique , using scientific coding method, dynamical scaler technique, efficient Cross-over strategy, we can optimize the speaker model, so that to increase recognition rate.iii) A new sort of network structure is put forward to settle the problem of the time during the NN speech recognition. It picks up the character vectors of a group of numbers from the sequence of the character vectors of the input speech signals, and then put the vectors into the NN sorter to recognize. Compared to the other methods of the NN speech recognition, with the network former transacting, it can shorten the time of training and recognition of the back-end NN sorter and simplify the network structure of the sorter to maintain higher recognition rate.This paper introduces the theory of the speaker recognition technique through above of three aspects. Finally, we introduce the realization of the speaker recognition system and the results of the speaker recognition experiments, and contrast with each other.