联合对角化能够成功解决盲分离问题,但在求解时会得到非期望的奇异解,从而无法完全分离出源信号.鉴于此,提出一种用于线性卷积混合盲分离的联合对角化方法,将卷积混合模型变换为瞬时模型,并对变换后的模型应用联合对角化求取分离矩阵.在求解过程中,引入约束条件对解的范围进行限定,避免了奇异解的出现.仿真结果表明,所提出的方法能够成功实现卷积混合信号盲分离. Joint diagonalization can successfully solve the problem of blind separation, but it can not get the complete separation of the source signal because of the unanticipated singular solution.In the view of this, a joint diagonalization method is proposed for the blind separation of linear convolution The method converts the convolutional mixed model into an instantaneous model and applies the joint diagonalization to the transformed model to obtain the separation matrix.In the solution process, the constraints are introduced to limit the scope of the solution, which avoids the emergence of singularity solutions. The simulation results show that the proposed method can successfully achieve convolutional mixed signal blind separation.