An effective approach for human activity recognition on smartphone

Activity recognition, which takes the sensor reading from mobile sensors as inputs and predicts a human motion activity using data mining and machine learning techniques. In this paper, we analyze the performance of two classification algorithm in an on-line activity recognition system working on Android platforms that supports on-line training and classification using only the accelerometer data. First we use the KNN classification algorithm and next we utilize an improvement of Minimum Distance and K-Nearest Neighbor classification algorithms, called Clustered KNN. For the purpose of on-line activity recognition, clustered KNN eliminates the computational complexity of KNN by creating clusters, i.e., creating smaller training sets for each activity and classification is performed based on these reduced sets. We evaluate the performance of these classifiers on four test subjects for activities of walking, running, sitting and standing in on-line activity recognition system. In this paper, we are also interested in the performance of classifiers with limited training data and the limited memory available on the phones compared to off-line.