基于窄带谱能量的快速正弦分析方法  

Fast sinusoidal analysis algorithm based on energy of narrowband spectrum

作　　者：杨萃[1] 韦岗[1] 

机构地区：[1]华南理工大学电子与信息学院,广州510640

出　　处：《声学学报》2009年第5期462-470,共9页Acta Acustica

基　　金：国家自然科学杰出青年基金(60625101);国家自然科学青年基金(60502041)资助项目

摘　　要：文中提出了一种基于窄带谱能量的快速正弦参数分析方法。该方法将信号η阶导数的傅氏变换与传统傅氏变换相结合,并且根据窗函数的主瓣能量集中特性,用原始信号的窄带能量逼近单个正弦分量的全带能量来提取正弦参数。实验表明,与已有算法相比,该算法的提取精度提高了30%～60%,计算量降低了4倍以上,可应用于实时语音分析和语音通信。A fast sinusoidal parameters extraction algorithm based on energy of narrowband spectrum is presented. To improve analysis accuracy and computational efficiency, the algorithm integrates the Fourier transforms of signal and its n-order signal derivatives. And energy of narrow band spectrum is used to approximate the energy of whole spectrum of single sinusoidal signal, since most energy is concentrated in the main lobe of spectral windows. Experiment results show that compared with existing algorithms, the proposed algorithm improves analysis precision by 30%-60% and reduces computational complexity, therefore it enables to use to real time signal analysis and speech communication.

关 键 词：谱能量 正弦 窄带 傅氏变换 N阶导数 能量集中 语音通信 语音分析 

分 类 号：O423[理学—声学]