基于Chirp原子分解的语音信号时频结构分析  

A Time-Frequency Analysis Method for Speech Signal Based on Chirp Atoms Decomposition

作　　者：武明勤[1] 于凤芹[1] 

机构地区：[1]江南大学通信与控制工程学院,江苏无锡214122

出　　处：《江南大学学报（自然科学版）》2006年第6期685-687,692,共4页Joural of Jiangnan University (Natural Science Edition)　

基　　金：江苏省自然科学基金资助项目(BK2002068)

摘　　要：从自适应信号分解的角度出发,使用匹配追踪算法将语音信号分解成非平稳Chirp原子的组合,根据单个Chirp原子的WVD分布线性组合分析语音信号的时频结构.仿真结果表明,该方法时频分辨率高,没有短时傅立叶变换中测不准原理的限制,不受传统WVD分布中有交叉项的影响.The paper presents a method in speech＇s Time-Frequency analysis to break the limit of Uncertainty Principle in STFT and to overcome the problem of cross-term interference in WVD analysis. This method is based on the principle of adaptive signal decomposition. Firstly, we use the algorithm of Matching Pursuit to decompose the speech signal into Chirp atoms, then, we analyse the speech＇s Time-Frequency structure by combining the decomposed Chirp atoms＇ WVD. The simulations to speech signal have shown that this method has high Time-Frequency resolution without the effect of cross-term interference in WVD analysis.

关 键 词：自适应信号分解 匹配追踪 魏格纳-维尔时频分布 短时傅立叶变换 

分 类 号：TN912.3[电子电信—通信与信息系统]