A comparative study on sign recognition using sEMG and inertial sensors

Abstract

Decision tree and random forest algorithm are introduced to the field of gesture recognition, and the gestures are classified by the fusion information of sEMG and inertial sensor. Experiments show that gesture recognition based on multi fusion information is more accurate than only using surface EMG or inertial signals. Taking the common 12 kinds of gestures as an example, the average recognition rate is 93.6%, some can even reach 100%. The cross validation indicate that the random forest is better than the decision tree in gesture recognition. Finally, the gesture recognition methods based on multi fusion information are used in the family service robot of our laboratory. By using hand gestures to control the service robot, a new human-computer interaction was explored.