摘要
独立分量分析(ICA)在频域中进行语音盲信号分离(BSS)时,将产生频谱分辨率降低和信号源间频谱相互干扰的矛盾,矛盾的任何一方突出时都会影响分离效果.为了解决这一矛盾,我们把短时综合的叠接相加法引进到BSS中,这一方法有效地缓解了这对矛盾,并且分离性能得到了明显的改善.仿真实验表明,这一方法简单可行并产生了很好的分离效果.
The contradiction of spectral resolution and permutation alignment will appear in speech blind signal separation(BSS) of frequency domain when using independent component analysis(ICA). Any side of the contradiction will degrade the separation performances. In order to solve the contradiction, authors introduce the short-time synthesis overlap-add method to BSS, which trades off effectively this contradiction and enhances evidently the separation performances. Authors validates the simplicity and feasibility of the method and shows very good separation results through the experiments of simulation.
出处
《四川大学学报（自然科学版）》
CAS
CSCD
北大核心
2003年第6期1071-1074,共4页
Journal of Sichuan University(Natural Science Edition)
关键词
语音盲信号
频谱分辨率
信号源间频谱
叠接相加法
speech blind signal separation
spectral resolution
permutation alignment
short-time synthesis overlap-add method
