摘要
盲源分离是近年来发展起来的一种新的信号分离技术,具有很好的应用前景.针对卷积混合模型,采用短时傅里叶变换将时域信号转化为频域,然后采用瞬时混合模型中的联合近似特征矩阵对角化算法,对频域带噪语音进行分离.从而实现语音降噪的目的,并解决了频域中存在的信号尺度和排序不确定性问题.通过在matlab上的仿真实验充分证明了该方法在语音降噪中的可行性.
Blind source separation, a new signal separation technique developed in recent years, has a good prospect of application. For convolutional mixture model, the study used the short-time Fourier transform of the time domain signal into a frequency domain, and then de-noised the noisy speech signals in frequency domain with the instantaneous mixture model algorithm of joint approximate diagonalization of eigenmatri- ces, and achieved the purpose of reducing the noise in voice and resolved the problem of the uncertainty of scale and sequencing in frequency domain signal. It fully proved the feasibility of the method in voice denoising by matlab simulation.
出处
《湖北工业大学学报》
2013年第1期75-77,共3页
Journal of Hubei University of Technology
关键词
盲源分离
卷积
傅里叶变换
降噪
blind source separation
convolution
fourier transform
de-noising
