EMG Based Decoding of Object Motion in Dexterous, In-Hand Manipulation Tasks

Abstract

Brain Machine Interfaces (BMI) are used to establish a communication pathway between the human brain and machines. Using BMI, signals from the brain are transmitted to an external processing unit where they are decoded into meaningful actions (e.g., browsing the internet using a PC or grasping an object with a prosthetic hand). BMIs are used to increase intuitiveness of the control of technical devices that can help individuals with motor or sensory impairments to regain their lost dexterity or able-bodied people to augment their capabilities. In this work, we present an Electromyography (EMG) based method for decoding object motion in dexterous, in-hand manipulation tasks. To do that, we use EMG signals derived from specific muscles of the human hand and forearm, and an optical motion capture system that records the object motion. The decoding is formulated as a regression problem using the Random Forests methodology that is based on a combination of decision trees. The model was trained using time-domain features, namely: root mean square, waveform length and zero crossings. A 5-fold cross validation procedure is used for model assessment purposes. This preliminary study achieves significantly high estimation accuracies, proving that object motion can be directly decoded from myoelectric activations of the muscles of the human hand and forearm. This work can support the formulation of EMG based telemanipulation schemes for advanced robotic and prosthetic hands.