基于脑肌电融合的混合脑机接口研究

动作模式识别是脑机接口技术的核心内容之一。针对目前脑机接口动作识别模式单一、识别率低等问题，基于混合脑机接口思想，提出一种脑电和肌电特征融合策略，可实现单侧肢体不同动作模式的有效分类，进而可用于脑机接口技术。同步采集9名健康受试者单侧手腕屈/伸两种动作模式下的脑电信号和表面肌电信号，分别提取脑电信号事件相关去同步化特征和表面肌电信号的积分肌电值特征，构建基于支持向量机和粒子群优化算法的脑肌电融合及运动模式识别模型，通过调整“特征融合系数”来实现动作模式最优分类，从而提高模式识别的准确率；进一步通过递降健康人的肌电信号幅值来模拟患者和运动疲劳状态下的肌电信号，验证所提出方法对动作模式识别的有效性。实验结果表明，基于脑肌电融合特征的动作模式识别率(98%)比单纯依靠脑电特征的识别率(73%)提高25%；在运动疲劳状态下，基于脑肌电融合特征的识别率稳定在80%以上，比单纯依靠肌电特征的识别率提高14%。可见，脑肌电融合策略能提高动作模式识别的准确性和鲁棒性，为混合脑机接口技术提供条件。

Multimodal Fusion of EEG and EMG Signals for a Hybrid BCI

Pattern recognition is one of the hot researches in the brain computer interface technology. In order to solve the problems in BCI, such as movement pattern singleness and low recognition rate, a hBCI based strategy fusioning the features of EEG and EMG was proposed to realize the classification of different motor patterns with unilateral limb. In the present study, the event related desynchronization features and intergrated electromyogram features were abstracted based on the EEG over brain scope and EMG from flexor carpi ulnaris and extensor carpi radialis longus under wrist flexion or extension in 9 healthy subjects. Secondly, the pattern recognition model fusioning the features of EEG and EMG, based on the theories of support vector machine and particle swarm optimization, was designed to classify optimally by adjusting the feature fusion coefficient. Furthermore, the proposed method was verified based on the EMG signals of patients or healthy subjects under fatigue state, which were simulated by descending the EMG amplitude of healthy subjects. Results showed that the recognition rate based on the fusion of EEG and EMG (98%) improved 25% compared to sole EEG feature under natural condition (73%); the recognition rate reached a stable level above 80% and improved 14% compared to sole EMG feature under fatigue state. It is revealed that the fusion of EEG and EMG feature contributed to improve the accuracy of pattern recognition and stability of movement, and provided the basis for the application of hybrid brain computer interface.