Investigation of Real-Time Control of Finger Movements Utilizing Surface EMG Signals

Surface electromyography (sEMG) has been the subject of investigation for the control of myoelectric prosthesis since the 1960s. Ongoing research seeks to improve existing systems as the complexity of the human hand and the challenge of obtaining meaningful control signals from the human body have created ample opportunities for further exploration. In the past, little focus has been placed on minimizing the number of sEMG channels required for pattern recognition of individual finger movements, as well as data preparation techniques used to optimize classification for such systems. The objective of this article is to describe the process required to obtain real-time sEMG classification while optimizing the number of individual finger movements made, as well as minimizing the number of sEMG channels required. Necessary data preparation and collection methods to optimize the system are also detailed. The resultant system classified four movements at an average accuracy of 72.2% in real-time, and made use of a multilayer perceptron (MLP) to achieve this. Due to the constraints imposed by the ethical clearance granted by the Research Ethics Committee (REC) of Stellenbosch University, the development of this system relied solely on data obtained from a single subject.