Blind Source Separation Of Binary Array Using ICA And Complex Neural Network

To reduce the delay of blind source separation (BSS) algorithm and improve its accuracy and stability, combining the advantages of traditional BSS technology and deep neural network, this paper proposes a binary array BSS technology based on independent component analysis (ICA) and complex neural network. A dual channel complex neural network based on time-frequency domain is proposed. Meanwhile the permutation problem in ICA is solved. The complex domain neural network is used to initialize the separation matrix based on the input mixed signal, the output of the neural network is in complex domain, the complex matrix is estimated by the learning label, and the target separation matrix is obtained by ICA. Compared with other independent component analysis methods, the proposed method improves the real-time performance and accuracy of blind source separation.