摘要
给出了一种基于小波变换和ＡＲＭＡ参数提取算法的声码器．该声码器在１．２ｋｂｉｔ／ｓ的速率上，重建语音的可懂度，自然度较好．该算法引入小波变换来实现子带滤波器组的设计，并引入多分辨分析的思想调整各子带的时窗宽度．各子带的声道模型由ＡＲＭＡ模型匹配，子带激励由冲激函数担当，故算法简单，易于实时实现．
A vocoder based on wavelet transform and ARMA parameters extraction is presented.The reconstructed speech is found at 1.2 kbit/s to be intelligible and relatively natural. WT(Wavelet Transform) is used in the filter bank design and an idea of multi resolutions analysis is exploited for adjusting the width of sub band time window.The model for vocal tract of sub band is replaced by ARMA model and the excitation of sub band is pulse function ,thus the algorithm is simple and more suitable for real time implementation.
出处
《北京邮电大学学报》
EI
CAS
CSCD
北大核心
1997年第4期1-5,共5页
Journal of Beijing University of Posts and Telecommunications
