| 语音MFCC特征计算的改进算法 |
| |
| 引用本文： | 章熙春,曹燕,张军,韦岗.语音MFCC特征计算的改进算法[J].数据采集与处理,2005,20(2):161-165. |
| |
| 作者姓名： | 章熙春 曹燕 张军 韦岗 |
| |
| 作者单位： | 华南理工大学电子与信息工程学院,广州,510640 |
| |
| 基金项目： | 国家自然科学基金(60172048)资助项目。 |
| |
| 摘 要： | 提出了一种计算Mel频倒谱参数(Mel frequency cepstral coefficient，MFCC)特征的改进算法，该算法采用了加权滤波器分析(Wrapped discrete Fourier transform，WDFT)技术来提高语音信号低频部分的频谱分辨率，使之更符合人类听觉系统的特性。同时还运用了加权滤波器分析(Weighted filter bank analysis，WFBA)技术，以提高MFCC的鲁棒性。对TIMIT连续语音数据库中DR1集的音素识别结果表明，本文提出的改进算法比传统MFCC算法具有更好的识别率。
|
| 关 键 词： | 语音识别 弯折离散傅里叶变换（WDFT） Mel频标倒谱参数 加权滤波器分析 |
| 文章编号： | 1004-9037(2005)02-0161-05 |
| 修稿时间： | 2004年4月11日 |
Improved Extraction Algorithm for MFCC Feature |
| |
| ZHANG Xi-chun,CAO Yan,ZHANG Jun,WEI Gang.Improved Extraction Algorithm for MFCC Feature[J].Journal of Data Acquisition & Processing,2005,20(2):161-165. |
| |
| Authors: | ZHANG Xi-chun CAO Yan ZHANG Jun WEI Gang |
| |
| Abstract: | An improved extraction algorithm for Mel frequency cepstral coefficient (MFCC) feature is presented. The new algorithm uses the wrapped discrete Fourier transform (WDFT) to improve the low frequency resolution of the input speech, which coincides with the characteristics of human auditory system. The weighted filter bank analysis (WFBA) is also used to improve the robustness of the MFCC. Phoneme recognition experiments on DR1 set of TIMIT continuous speech corpus show that the proposed algorithm can effectively improve the recognition rates. |
| |
| Keywords: | speech recognition wrapped discrete Fourier transform (WDFT) Mel frequency cepstral coefficient (MFCC) weighted filter bank analysis (WFBA) |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
