摘要
MFCC可以更加准确地表示语音低频处的频谱包络,在语音转换中是一种很好的频谱包络表示方法;提出一种新的有效的求解MFCC的算法,先对频谱通过谐波抽取得到阶梯谱包络,再对阶梯谱包络进行Mel尺度变换,最后通过余弦变换求得MFCC参数。该方法实现简单,运算结果准确。
MFCC can express speech spectrum more accurately at low frequency, so MFCC is a good method to express the spectral envelope in voice conversion; A new and efficient algorithm for MFCC is proposed. At first, the staircase envelope is obtained through harmonic extraction, then the staircase envelope is transformed by Mel frequency scale, at last the MFCC coefficients are extracted by cosine transformation. The algorithm can be realized easily, and the result is good.
出处
《国防科技大学学报》
EI
CAS
CSCD
北大核心
2004年第4期42-45,共4页
Journal of National University of Defense Technology
关键词
语音处理
MFCC
语音识别
说话人识别
语音转换
speech processing
MFCC
speech recognition
speaker recognition
voice conversion