Human Activity Recognition using Android Smartphone

Activity recognition is one of the most important technology behind many applications such as medical research, human survey system and it is an active research topic in health care and smart homes. Smart phones are equipped with various built-in sensing platforms like accelerometer, gyroscope, GPS, compass sensor and barometer, we can design a system to capture the state of the user. Activity recognition system takes the raw sensor reading from mobile sensors as inputs and estimates a human motion activity using data mining and machine learning techniques. In this paper, we analyze the performance of two classification algorithms i.e. KNN and Clustered KNN in an online activity recognition system working on Android platforms and this system will supports on-line training and classification using the accelerometer data only. Usually first we use the KNN classification algorithm and next we utilize an improvement of Minimum Distance and K-Nearest Neighbor classification algorithms, i.e. Clustered KNN . For the purpose of activity recognition, clustered KNN will eliminates the computational complexities of KNN by creating clusters (creating smaller training sets for each actions and classification will be performed based on these reduced training sets). We can predict the performance of these classifiers from a series of observations on human activities like walking, running, lying down, sitting and standing in an online activity recognition system. In this paper, we are intended to analyze the performance of classifiers with limited training data and limited accessible memory on the phones compared to off-line.