论文部分内容阅读
为实现卷积混合信号的盲分离,提出了一种基于斜投影的子空间方法,首先设计“过去”、“现在”和“未来”的观测数据空间,并通过斜投影将卷积混合转化成为线性瞬时混合;然后采用静态分离算法重构源信号。该方法利用了观测数据矩阵的结构信息直接获得线性瞬时混合的数据模型,不需要进行高维子空间代价函数的优化,运算量相对小。计算机仿真验证了算法的有效性。
In order to realize the blind separation of convolutional mixed signals, a subspace method based on oblique projection is proposed. First, the space of observation data of “past”, “now” and “future” are designed and analyzed by oblique projection The convolutional mixture is transformed into a linear instantaneous mixture, and then the static separation algorithm is used to reconstruct the source signal. The method utilizes the structure information of the observed data matrix to directly obtain the linear instantaneous mixed data model, and does not need to optimize the cost function of the high dimension subspace, and the calculation amount is relatively small. Computer simulation verifies the effectiveness of the algorithm.