Perturbation Analysis of Mel-Frequency Cepstrum Coefficients
Mel-frequency cepstrum coefficient (MFCC) is a widely used feature vector in speech signal precessing. Its feature extraction procedure can be seen as a mapping function which transfers the input speech signals to output MFCC feature vectors. However, this function is too complex to analyze and even a simple approximation is not easy to obtain. This paper studies the effects of each MFCC feature extraction step and obtains the relation between the input signal-to-noise ratio (SNR) and the output perturbation bound of MFCC feature vectors. Experimental results show that the obtained bounds are tight and nearly full covered. This analysis method may help us to find new clue of MFCC and may has potential applications in speech recognition.
Wei-Qiang Zhang Dengzhou Yang Jia Liu Xiuguo Bao
Tsinghua National Laboratory for Information Science and Technology, Department of Electronic Engine Institute of Electronics, Chinese Academy of Sciences, Beijing 100190, China CNCERT/CC, Yumin Road, Chaoyang District, Beijing 100029, China
国际会议
上海
英文
715-718
2010-10-20（万方平台首次上网日期，不代表论文的发表时间）