Abstract
A high-efficiency, easy-to-use input device is not only important for data entry but also for human-computer interaction. To date, there has been little research on input devices with many degrees of freedom (DOF) that can be used by the handicapped. This paper presents the development of an electromyography (EMG)-based input device for forearm amputees. To overcome the difficulties in analysing EMG and realising high DOF from biosignals, the following were integrated: (1) an online learning method to cope with nonlinearity and the individual difference of EMG signals; (2) a smoothing algorithm to deal with noisy recognition results and transition states; and (3) a modified Huffman coding algorithm to generate the optimal code, taking expected error and input efficiency into consideration. Experiments showed the validity of the system and the possibility for development of a quiet, free-posture (no postural restriction) input device with many DOF for users, including forearm amputees.