与传统的鼠标、键盘等人机交互方式相比，表面肌电信号（surface Electromyography, sEMG）作为一种伴随运动产生的生物电信号可以提供一种更加自然直观的人机交互方式，具有广泛的应用价值。本项目从sEMG信号的特点出发，针对sEMG存在个体差异性和概念漂移的问题，拟通过设计和改进相应的特征提取和模式识别算法，提高肌电动作识别系统的分类准确率和泛化能力。针对多通道电极信号存在多重共线性问题，采用最小二乘回归算法进行非线性回归，实现基于sEMG的手指关节连续运动的估计。主要研究内容包括：1) 受试者无关肌电信号特征提取与肌电信号“概念漂移”检测算法；2) 自适应增强肌电动作分类器构建算法研究; 3) 基于多模态信息融合的肌电手势动作识别算法研究；4) 基于PLSR的连续手指运动估计算法研究。

As bioelectricity which is generated along with movement, sEMG is a more natural and intuitive control media when compared with traditional human-computer interface, such as mouse, and keyboard. It has wide applications. In this project, we will start with handling the disadvantages of sEMG. By designing and improving corresponding feature extraction methods and pattern recognition methods, we are going to eliminate the adverse effect of individual difference and concept drift in sEMG on finger gesture recognition. For the multiple conlinearity prolems in multiple sEMGs, we are going to apply PLSR for regression. The main research contents include: 1)Algorithm on user-indepdendent sEMG feature extraction and sEMG "concept drift" detection; 2) Finger movement recognition by using adaptive self-enhancing classifier 3) Finger movement recognition based on multimodal information fusion; 4)Continous finger joint angle estimation by using sEMG with PLSR.