欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
将基于协方差矩阵对角化的盲分离方法用于盲分离。当信号具有时间结构且空间独立时,通过对角化独立成分的自协方差矩阵（不同时延下的协方差矩阵）对混合信号进行盲分离。首先构造白化信号的时延协方差矩阵,然后选取不同的时延,将协方差矩阵的对角化程度表示成代价函数,最后利用梯度下降法得到分离矩阵。此方法和基于极大似然估计的FastICA算法的对比试验说明了此算法的有效性。
英文摘要：
Algorithm based on diagonalization of the covariance matrix was used to separate speech signals in this paper. We can separate the mixtured signals by diagonalizing the covariance matrix in different time - lag when they have the temporal structure and are independent in space. Firstly, we construct the covariance matrix of the whited signals, and then choose different time-lags so as to construct the cost-function. At last we find the separating - matrix by using the method of grad - descending. Computer simulations based on this method and the MLfastica are presented to show the effectiveness of the proposed algorithm.
同期刊论文项目
同项目期刊论文
An efficient and fast algorithm for estimating the frequencies of superimposed exponential signals i
Unified parametric and nonparametric ICA algorithms for hybrid source signals and stability analysis
期刊信息
位置：