Towards an EMG Controlled Prosthetic Hand Using a 3D Electromagnetic Positioning System

This paper proposes a novel method of using electromyographic (EMG) potentials generated by the forearm muscles during hand and finger movements to control an artificial prosthetic hand worn by an amputee. Surface EMG sensors were used to record the forearm EMG potential signal via a PC sound card, meanwhile, a novel 3D electromagnetic positioning system together with a data-glove mounted with 11 miniature electromagnetic sensors was used to acquire human hand motion in real time. The synchronized measurement of hand posture and EMG signal are stored as prototypes, in the format of a series of data frames, each comprising a set of positional and orientation posture data and a set of EMG data. A graphical hand model was also generated to visualize the real time hand movement. Further, we propose that only the EMG measurement device be attached to the forearm muscle of the prosthetic hand user. Candidate sets of EMG data acquired in real time will be compared with stored prototypes within each data frame using a pattern recognition approach. Subsequently, the most likely posture data set in this frame will be considered as the numerical expression of the current hand shape and used to control a robotic hand so that it carries out the user's desire. With a two-channel EMG measurement device, we first apply frequency analysis on the conditioned raw EMG signal. Then, pattern recognition techniques may be applied to identify the most closely aligned spectrum generated from the data recorded by the dual channel EMG measurement device. Alternatively, when a multi-channel (5-6) EMG measurement device is developed, pattern recognition can be applied on the amplitude of EMG signals to identify the most likely EMG signal distributions. This approach offers several advantages over existing methods. Firstly, it will simplify the classification procedure, saving computational time, secondly, it will reduce the requirement for the optimization process, and finally it will increase the number of recognizable hand shapes and subsequently improve the dexterity of the prosthetic hand and the quality of life for amputees. The database of EMG prototypes could be employed to optimize the accuracy of the system within a machine learning paradigm. By making a range of EMG prototype databases available prosthetic hand users could train themselves to use their prosthesis using the visual reference of the real time hand model to provide feedback