A Low Cost Embedded Solution for Detection and Analysis of Three-Dimensional Gestures through Wireless Communication

Abstract This paper describes the concept of a non-optical acceleration-based gesture recognition system for a low cost and small size remote control for mobile robots which is capable of recognizing more than 5 gestures based on the input signals from a MEMS 3-axes accelerometer and a MEMS 2 axes-gyroscope. The system is designed like a wristwatch and based on a microcontroller module which detects the accelerations and the movements of the user's arm and transmits the results wirelessly to a receiver. The two main topics in this paper are the description of the hardware and the newly developed algorithm used to analyze and classify the detected movements in comparison to the traditional algorithms Hidden Markov Model (HMM) and Dynamic Time Warping (DTW).