Signal reconstruction of compressed sensing based on recurrent neural networks

In this paper, neural network approach is addressed for signal reconstruction under the frame of compressed sensing. By introducing implicit variables, we convert the basis pursuit denoising model into a quadratic programming problem. Based on a class of generalized Fischer–Burmeister complementarity functions, we establish a neural network method for the signal reconstruction of compressed sensing. A projection neural network is also presented to recover the original signals. These two neural networks can be implemented using integrated circuits and two block diagrams of the neural networks are presented. Based on our proposed method, some potential applications of the compressed sensing are discussed.