Head gesture recognition via dynamic time warping and threshold optimization

Gesture recognition is one of the emerging fields in industry and a hot research topic in academia. It is commonly used in smart devices to assist the owners in their day-to-day life. But it is also important in facilitating processes in any kind, that involves people. In our attempt at improving life quality for disabled people below the neck, an assistive autonomous powerchair is developed. To ease interaction with the chair, we propose embedding a head gesture recognition system using an IMU (Inertial Measurement Unit) sensor. This study explores the possibilities of such implementation. Several approaches have been developed for gesture recognition. Accuracy, sensitivity and rapid computation are some of the critical items which are being considered in different approaches. In this study, we use the Dynamic Time Warping (DTW) algorithm in order to calculate the similarity between two time sequences. After DTW calculation, we propose a new approach which optimizes the decision making problem and calculates the optimum threshold values. We propose and compare two different simple geometrical shapes for threshold optimization. Even with these simple 3D objects, 85.68% success rate is achieved. This means that more than 8 out of 10 repetitions of a gesture are recognized successfully. The results are promising for future studies.