基于神经动力学优化的压缩感知信号恢复方法

针对稀疏信号的准确和实时恢复问题，提出了一种基于神经动力学优化的压缩感知信号恢复方法。通过引入反馈神经网络（Recurrent Neural Network, RNN）模型求解l1范数最小化优化问题，计算RNN的稳态解以恢复稀疏信号。对不同方法的测试结果表明，提出的方法在恢复稀疏信号时所需的观测点数最少，并且可推广到压缩图像的恢复应用中，获得了更高的信噪比。RNN模型也适合并行实现，通过GPU并行计算获得了超过百倍的加速比。与传统的方法相比，所提出的方法不仅能够更加准确地恢复信号，并具有更强的实时处理能力。

A Neurodynamic Optimization Method for Recovery of Compressive Sensed Signals

Aiming at the problem of accurate and real-time recovery for sparse signals, this paper developed a neurodynamic optimization method to reconstruct compressive sensed signals. By introducing recurrent neural network (RNN) to solve the l1 norm minimization problem, the proposed method could recover sparse signals by computing stable solution of the RNN. Results of tests for different methods show that the proposed method requires minimum measurement points to recover sparse signal, and can be applied for recovery of compression image to obtain a higher signal to noise ratio. The RNN model is also suitable for parallell implementation, and obtains more than 100 times speedup by GPU parallel computing. As compared with the conventional methods, the proposed method can not only recover signals more accurately, but also hold a better real-time processing capability.