本文研究如何分离具有延迟而无回波的混叠信号。提出了欠定情况(即感知器少于源信号)下具有时间延迟的信道估计算法,称为单源区间衰减和延迟恢复(AD-MRISSI)算法;它在理论上能无偏差地估计信道,在应用中能准确地估计延迟和精确地估计衰减。也提出了一个基于稀疏原则的源信号重建算法,能避免Bofill算法中的求解二阶锥规划问题。几个语音信号实验展示了其有效性和实用性.
How to separate delayed mixtures without any reverberation is studied.An algorithm to estimate channel with delays for the underdetermined model (i. e., fewer sensors than sources) is proposed. It is called attenuation and delay matrix recovery in single source intervals (AD-MRISSI) algorithm. Theoretically, it can estimate the channel without bias. Practically, it can estimate the delays and the attenuations exactly.An algorithm to recover the sources is also proposed based on a sparse criterion. It can avoid solving second-order cone programming problem in Bofill' s algorithm. Several experiments of sound sources demonstrate the effectiveness and practicality of our algorithm.