首页>
外国专利>
A Speaker Identification Method Based on DTW Algorithm
A Speaker Identification Method Based on DTW Algorithm
展开▼
机译:基于DTW算法的说话人识别方法
展开▼
页面导航
摘要
著录项
相似文献
摘要
#$%^&*AU2018102038A420190131.pdf#####ABSTRACT This invention is a speaker-recognition method, in speech recognition technology field, based on Dynamic Time Warping (DTW) Algorithm. The whole invention consists of several steps: conduct a pre-process of the speakers' auditory signals using normalization method, effective auditory segment detection and the Mel Frequency Cepstral Coefficients(MFCC) feature extraction, etc. ; divide the processed auditory signals into training sets and test sets; measure the distances between the amplitudes of each two factors in the two different sets using DTW Algorithm, and then, the one whose auditory signal distance is found the most frequently in the first K shortest distances would be identified as the test speaker, so that finalize the identification of the speech speaker. This invention does not require tester's participation and adjustment and can automatically perform feature extraction and recognition functions, which provides a reliable and fast DTW algorithm-based speaker identification technology.Fig. 5 Process diagram of MFCC arithmetic 25 . ,, , , FMCC 25: 20 Kl ||| n 15-5 -111 -00 100 00 300 00 500 600 700 800 frame number Fig. 6 The signal features extracted by MFCC
展开▼