Real-Time Motion Detection for Android Smartphones

Real-time motion detection can be used to monitor a wide range of physical activities, such as those typically leveraged by exergames, but the literature present few algorithms. Motivated by this demand, we aimed at evaluating and proposing motion detection algorithms in Android smartphones in order to comply with three requirements: accuracy, precision, and recall above 95%. We evaluated the existing algorithms, Google and Step Detector. The results of this evaluation showed the Google algorithm as the most promising, but it did not meet all the requirements. Therefore, a new algorithm, based on it, was proposed and evaluated. Thus, this paper introduces the Castor algorithm and a tool for analyzing motion detection algorithms. In our evaluation, Castor was able to overcome Google in most scenarios and could adapt to lower-speed movements, when Google's algorithm could not. As contributions, this paper presents Castor as an algorithm to detect movements and also a tool which can be extended for implementing and evaluating new algorithms.