Extracting signatures from muscle activity data for control of myolectric prosthesis

Upper limb myoelectric prostheses are controlled by the voluntary contraction of residual muscles of the amputated limb. These contractions produce weak electric signals that can be captured using surface electromyography (sEMG) which can be used to control the prosthesis. A number of multi-articulate prostheses have recently been developed that utilise sEMG control, offering the potential for increased usability. The intent is that the patient will be able to reproduce the same muscle contractions that they would have normally without the prosthesis in order for it to perform the associated hand gesture or finger movement [7]. In this project, we explored and investigated the existence of sEMG signatures associated with particular hand gestures and finger movements. Experiments were designed to record sEMG activity from healthy subjects. Our experiments used four sensors instead of the two that the current model uses, due to limitations in pattern recognition with only two sensors. We looked into signal processing and pattern recognition methods to find if there is any commonality amongst the different subjects so that the iLIMB could identify the gestures that someone makes without having to program it for each individual [14]. The features we found showed promise for future developments in pattern recognition of myoelectric signals. The multi-articulate prosthesis we used was the Touch Bionics iLIMB. This is an advanced myoelectric prosthesis where each digit is individually powered, allowing the prosthesis to be positioned into several gestures. The current model uses two electrodes, one placed on the extensor muscle and the other on the flexor muscle for EMG control.Upper limb myoelectric prostheses are controlled by the voluntary contraction of residual muscles of the amputated limb. These contractions produce weak electric signals that can be captured using surface electromyography (sEMG) which can be used to control the prosthesis. A number of multi-articulate prostheses have recently been developed that utilise sEMG control, offering the potential for increased usability. The intent is that the patient will be able to reproduce the same muscle contractions that they would have normally without the prosthesis in order for it to perform the associated hand gesture or finger movement [7]. In this project, we explored and investigated the existence of sEMG signatures associated with particular hand gestures and finger movements. Experiments were designed to record sEMG activity from healthy subjects. Our experiments used four sensors instead of the two that the current model uses, due to limitations in pattern recognition with only two sensors. We looked into signal processing and pattern recognition methods to find if there is any commonality amongst the different subjects so that the iLIMB could identify the gestures that someone makes without having to program it for each individual [14]. The features we found showed promise for future developments in pattern recognition of myoelectric signals. The multi-articulate prosthesis we used was the Touch Bionics iLIMB. This is an advanced myoelectric prosthesis where each digit is individually powered, allowing the prosthesis to be positioned into several gestures. The current model uses two electrodes, one placed on the extensor muscle and the other on the flexor muscle for EMG control