摘要
盲信号处理算法主要有批处理算法和自适应算法两类,导出了一种批处理和自适应相结合的快速独立分量分析(fastindependent component analysis,Fast ICA)算法。将该算法应用于语音信号盲分离处理,通过综合实验,从分离前后的波形、频谱图和主要评价参数说明该算法具有良好的信号分离效果。与扩展联合对角化(jointapproximativediagonalizationof eigenmatrix,JADE)算法和自然梯度(natural gradient,NG)算法比较,Fast ICA算法具有更好的分离效果。
The main types of blind signal processing algorithm are batch algorithm and adaptive algorithm.Combined with batch algorithm and adaptive algorithm,the fast independent component analysis algorithm for speech signal blind separation processing is presented.Through the comprehensive experiments,the results show that Fast ICA algorithm has good signal separation efficiency from the signal waveforms and spectrums before and after separation and the main evaluation parameters.Fast ICA algorithm has better separation efficiency than the joint approximative diagonalization of eigenmatrix algorithm and natural gradient algorithm.
出处
《计算机工程与设计》
CSCD
北大核心
2010年第13期3047-3050,共4页
Computer Engineering and Design
基金
广东省教育厅育苗工程基金项目(粤财教[2008]342号)
广东省自然科学基金项目(07010869)
关键词
盲信号处理
语音信号盲分离
快速独立分量分析
批处理
自适应
blind signal processing
speech signal blind separation
fast independent component analysis
batch
adaptive
