A Prosthetic Hand Control System Based On SEMG And NIRS Signals Decoding

Near-infrared spectroscopy(NIRS) is a non-invaded technique to monitor human tissue oxygen concentration changes, which is more and more widely used in monitoring muscle activities with its high spacial resolution. Surface electromyography(sEMG) is the electrical activity produced by skeletal muscle and is recorded from the surface above the muscle on the skin. sEMG is a direct source that reflects the muscle activity and is widely used in human computer interface(HCI) nowadays,especially to control equipments to help the disabled in rehabilitation engineering, like a prosthetic hand or a wheelchair. However, HCI based on sEMG is easily influenced by noise, has low spacial resolution and will drift with the subject’s physiological condition(like muscle fatigue).Research has shown that, by introducing NIRS signal, the performance of the sEMG based HCI can be improved.In this paper, NIRS signal is introduced to monitor muscle group activity, and a prosthetic hand control system based on combined s EMG and NIRS signals decoding is developed. Experiments are carried out to validate the improvement by introducing NIRS. The main contents include:Firstly, a NIRS sensor for muscle activity monitoring is developed,and a series of tests show that noise, stability and anti-interference ability of the sensor are comparable the existed sensors, and the sensor can monitor muscle oxygenation change efficiently. Then, combined with the s EMG sensor previously developed in the lab, a fusion sEMG and NIRS sensor is developed by reducing the interference between s EMG and NIRS.Secondly, based on the sEMG acquisition software previously developed in the lab, a PC software for sEMG and NIRS featureextraction and pattern recognition is developed. Then, the pattern recognition result is used to control a under-actuated prosthetic hand previously developed in the lab to perform typical motion classes and a whole prosthetic hand control system based on combined s EMG and NIRS signals is developed.Finally, off-line pattern recognition experiments and on-line prosthetic hand control experiments are carried out. The results of off-line experiments show that, by introducing NIRS signal, the accuracy of off-line pattern recognition can be improved by 10%. The results of on-line experiments show that the prosthesis control system, by decoding s EMG and NIRS signals, can control the prosthesis to finish typical motion classes in real-time.