EMG Based MIDI Controller

MIDI is a standard that has been used for musical performance since the early eighties and has been gaining increasing popularity ever since for its ease of use and versatility. The objective for this work is to implement a MIDI controller that takes input from the user’s hand/arm motion, and as such, send MIDI data to a MIDI interface. Hand movement was captured using three EMG sensors (Olimex Shield-EKG-EMG) along the forearm region to capture muscle activity. These will be connected to an Arduino UNO board and send sensor data into a Raspberry PI 4B board for data analysis. Then, in order to compare and determine which one is best to detect four different hand gestures, said activity will be analyzed using three different machine learning algorithms, namely Random Forest (RF), Support Vector Machines (SVM) and a three hidden layer neural network (NN). After classification, depending on the gesture, the controller converts these movements into their correspondent MIDI note for musical performance. Finally, performance will be tested using virtual instruments on a digital audio work station (DAW), namely Ableton Live 9.