Human Activity Recognition System Including Smartphone Position

The data gathered by acceleration sensors in smartphones gives different results depending on the location of the smartphone. In this paper, a human activity recognition system was proposed, including the smartphone's position. This system can recognize not only the activity of a person, but also the location of the smartphone. HOG (Histograms of Oriented Gradients) were used to extract features of the acceleration data, because the waveform of the acceleration data is very complex. Then, a strong classifier was obtained using a learning algorithm of Real AdaBoost based on the position of possession smartphone and acceleration sensor data. It also improves the recognition rate by analyzing the acceleration data. The effectiveness of the activity recognition system was shown by the experiment.