植入式脑机接口是将电极植入人或动物的大脑皮层，获得神经元发放模式的信息，并且利用该信息来修复人或动物的感觉、运动等功能。目前，该研究已经成为生命科学的研究热点，并且取得了重大突破，但是在神经解码方面的研究仍面临巨大的挑战。本项目拟以大鼠为研究对象，研究基于运动皮层神经元锋电位序列的解码方法。针对现有解码方法受限于线性、静态的假设，建立基于广义回归神经网络的解码方法，实现非线性、动态的神经解码，提高传统线性、静态的解码效率；针对目前的单层解码策略，提出两层神经解码策略，并由两种较易实现的方法来完成，降低解码过程的计算消耗，并能够同时解码离散和连续运动；为了将各个神经元对运动信息的调制能力都达到最大化，本项目提出基于精细时移的神经解码策略，进一步改进解码模型的预测效果和效率，通过本项目的研究，期望获得高效、新颖的神经解码策略，为深入了解大脑的信息编码和运动控制机制奠定基础。

The implantable brain computer interface implanted electrodes in thecerebral cortex of human or animal to get the neural firing information, and used this information to restore the sensory and motor functions of the human or animal. At present, the study has become research focus of information field, and has made major breakthrough. However,it still faces huge challenge in neural decoding. In this project, rat's forelimb movement is studied, and decoding algorithms based on spike trains are proposed. Limited by linear and static assumptions, the existing method for decoding doesn't perform well. Neural decoding method based on general regression neural network is established to achieve non-linear, dynamic neural decoding, and improve the efficiency of traditional linear, static decoding method; two-stage decoding strategy is proposed and achieved by two easy methods.This strategy is aimed to reduce the calculation consumption in the decoding process, and decode discrete and continous movement simultaneously; neural decoding method based on precise time shift is proposed to further improve the decoding efficiency. This method enlarges neuron's modulation capabilities of the movement information. Efficient and novel neural decoding strategies will be realized through this research, and this will be useful for understanding the brain information coding mechanisms and movement control basis.