基于样本熵与时频分析的手部抓握运动意图预测

针对上肢外骨骼控制信号产生与外部设备响应存在时间滞后导致脑机接口(BCI)系统实时性差的问题，采集被试手部自主抓握前运动相关皮质电位(MRCP) ，提出基于非线性复杂度特征样本熵 (SampEn)与线性幅值特征融合算法的手部运动意图预测方法. 从时频、神经复杂度分析不同大脑状态之间存在的差异，通过特征融合实现对手部抓握运动意图的预测. 基于特征融合意图离线预测准确率最高可达88.46%，可以在人体手部自主运动发生时刻?1 400 ms实现对手部运动预测. 与平静时期手部静止状态相比，被试产生手部抓握运动意图时脑电信号的功率谱与复杂度均产生明显变化，为基于手部运动意图预测提前驱动机器人实现人机协同提供控制策略.

Prediction of hand grasping intention based on sample entropy and time-frequency analysis

The movement related cortical potentials (MRCP) before the autonomous grasping movement of the hand were collected, and based on the fusion algorithm of nonlinear sample entropy (SampEn) feature and linear amplitude feature a prediction method of motion intention was proposed, in order to solve the problem of poor real-time of brain computer interface (BCI) system caused by the time lag between the control signal of upper limb exoskeleton and the response of external devices. The differences between different brain states were analyzed from the perspectives of time-frequency and neural complexity, and the prediction of hand grasping motor intention by feature fusion was realized. The off-line classification accuracy of intention based on the feature fusion was up to 88.46%, when the occurrence time of human hand voluntary motion was ?1400 ms the prediction of hand motion can be realized. The EEG signals of hand grasping intention in the power spectrum and complexity are significantly different from those in the static state of the hand during the quiet period which could be proposed as a control strategy based on hand motion intention to drive robot in advance to realize human-computer cooperation.