This project focuses on the development of robot-assisted stroke rehabilitation by implementing electromyography (EMG) as the interface between robot and user communication. Key issue in implementation of EMG in this application is conversion of EMG signal into torque data. This paper presents a methodology of EMG signal to estimated joint torque conversion by using genetic algorithm (GA). Basic principle of GA, formulation, and implementation to the problem are discussed in this paper. Experimentation with real life EMG data has been carried out to assess the feasibility of the methodology in robot-assisted stroke rehabilitation problem. Preliminary investigations show that the methodology can be used in EMG to joint torque conversion algorithm.