Single Channel Blind Source Separation Under Deep Recurrent Neural Network

Abstract

In wireless sensor networks, the signals received by sensors are usually complex nonlinear single-channel mixed signals. In practical applications, it is necessary to separate the useful signals from the complex nonlinear mixed signals. However, the traditional array signal blind source separation algorithms are difficult to separate the nonlinear signals effectively. Building upon the traditional recurrent neural network, we improved the network structure, and further proposed the three layers deep recurrent neural networks to realize single channel blind source separation of nonlinear mixed signals. The experiments and simulation were conducted to verify the performance of this method; the results showed that the mixed signals can be separated excellently and the correlation coefficient can be reached up to 99%. Thus, a new method was given for blind signal processing with artificial intelligence.