Study On Speaker Verification Technology Related To Text And Applications

Speaker recognition is the process of automatically recognizing who is speaking on the basis of individual information include in speech signals. It has well application prospects in many fields. Speaker recognition methods can be divided into text-dependent and text-independent methods according to the verbal content of the speech signal. The thesis bases on dynamic time warping and Gaussian mixture model, studies endpoint detection, speaker feature extraction, training methods of the speaker model, and the noise robustness in-depth.The thesis find the signal's endpoint and filter the speech silence segment. We give comparison of the two endpoint examination methods: double-gate thresh-hold method and energy-frequency-value method. Experiments show that latter can partition the endpoint of noise speech better.The robustness of every components of the MEL Frequency Cepstral Coefficient for Gaussian additive background noise is studied. The optimal range of robust components is also evaluated. Combined with dynamic coefficients, promising results are given in some experiments.The thesis implements two methods of Spectral Subtraction and Cepstrum mean subtraction. The thesis implements text-dependent speaker verification system basing on Gaussian mixture model in the noise environment.The thesis implements a speaker verification system basing on system on programmable chip and RF smart card technology. The thesis discusses the fine prospect about the combination of system on programmable chip and RF smart card technology. The thesis introduces the method of dynamic time warping and the implement method.