SVM Classification for Novel Time Domain IMU and EMG fused features for control of 6-DOF industrial robot

Gesture recognition is an up and coming field with applications in the field of biomedical engineering, human computer interaction and other fields. Electro myogram sensors (EMG) and inertial measurement units (IMU) are often used to combine the vital information necessary for gesture recognition. This study provides a novel method to access the time domain features of IMU sensors and then fuses this information with the time domain features of the EMG sensors. Although various classification techniques are used to this end. This research uses the gesture recognized to control a virtual robot. This study presents the design of sensory system and collection of data. This study also deals with calculation of features for both EMG and IMU time series. This study visualizes the class separability using various visualization tools. The following classification methods are applied on these features, support vector machines (SVM)The results of these different methods are compared based on accuracy, precision, recall, f1-scores and ROC curves and area under ROC curves for each class of gestures. Finally, a JACO robot is controlled using the gestures in a virtual environment.