Notes about accelerometer-based gesture recognition problem

N. V. Kurilenko

Lomonosov Moscow State University, Faculty of Mechanics and Mathematics

Abstract: A solution of accelerometer-based gesture recognition problem has a variety of applications in robotics, mobile device manufacturing and mobile applications. It is a standard classification problem and in many cases it can be solved with relatively simple analysis of an output of sensors: accelerometer, gyroscope, magnetometer, etc. However, in cases where outputs for different gestures are visually in distinguishable, it is necessary to use advanced machine learning techniques. The paper examines one of these problems and high lights the topic of formulating the methodology of finding solutions for such problems.

Keywords: machine learning, data mining, gesture recognition, Hidden Markov Models, accelerometer.