基于简化模型的卷积混叠盲分离并行算法

电信会议所用的传感器往往相距较近。根据从同一源到两个靠得很近的麦克风的房间冲激响应具有相似性这一特点,把多输入多输出盲卷积问题简化为若干个单输入单输出卷积混叠盲分离问题。提出了一种新的使用时间预测性的多输入-多输出盲解卷算法,该算法通过简化模型和并行算法,能用广义特征值的方法直接求出解卷滤波器,因而算法复杂度小。使用音频信号的计算机仿真结果显示了算法的有效性。

Blind deconvolution parallel algorithm based on a simplified mixing model

The distance between sensors is usually closely in telecom meeting.MIMO can be simplified to SISO according as there is similar impact response when microphones are very nearly each other.A new kind of multi-input multi-output blind deconvolution using a measure of temporal predictability is presented.The algorithm can use an eigenvalue problem to solve directly a deconvolution filter by a simplified mixing model and parallel algrithm.Thus the complexity of the algorithm is small.The results of computer simulations are presented for sound signals.