介绍了独立分量分析的原理、算法及其特点。考虑到在语音分离中,声音信号复杂多样,但是不同语音源信号保持相对独立的特点,利用盲信号分离的思想,使用改进的FastICA方法用于语音信号分离,以获得独立的声音文件。文中使用两个声音文件做实验,取得了比较满意的效果。
The principle and algorithms of independent component analysis(ICA) was illustrated. Considering the diversity of the sound, which kept only comparatively independence in the source signal and using the idea of blind signal separating, the FastICA algorithms in sound separation was tried. Finally, two sounds were tested and some good results were got.