Research And Speaker Recognition Controller

Speech Recognition, which belongs to biometrics recognition technology, is a technique that recognizes speakers by scientific calculation on the basis of extracting feature from speech. One significant use of speaker recognition is to determine whether a speaker has the right to enter security or confidential systems. Using speech password has advantages the traditional way by inputting password on keyboard doesn't have, for it is unforgettable and cannot be easily taken. However, systems decline badly in performance because of the intra-speaker variability of a word. On the basis of studies on the feature extraction and recognition methods, this paper is focus on the study of Linear Prediction Cepstrum(LPCC) and One-Pass Dynamic Time Warping(One-Pass DTW). Also, a speaker verification system based on LPCC and One-Pass DTW is implemented on the Sunplus 16-bit MCU SPCE061A.This paper's major works are as follow:1. Some pre-processing technologies, including pre-filtering, pre-emphasis, enframe and endpoint detection, are introduced.2. The principle of human voice produced and digital model are introduced. The principles and extraction methods of Linear Prediction Cepstrum Coefficients (LPCC) and Mel-scaled Frequency Cepstrum Coefficients (MFCC) are introduced.3. Several methods of speaker recognition, especially Dynamic Time Warping (DTW) are studied. The method One-Pass DTW is deeply studied.4. Some issues of the speaker verification system implementation on the Sunplus 16-bit MCU SPCE061A are discussed. A speaker verification system is implemented on the SPCE061A platform.