Abstract

Recognition of human activity has a wide range of applications in medical research and human survey systems. We present a powerful activity recognition system based on a Smartphone in this paper. The system collects time series signals with a 3- dimensional Smartphone accelerometer as the only sensor, from which 31 features in the time domain and frequency domain are created. The quadratic classifier, k-nearest neighbor algorithm, support vector machine, and artificial neural networks are used to classify activities. Feature extraction and subset selection are used to reduce dimensionality. In addition to passive learning, we use active learning techniques to lower the cost of data tagging. The findings of the experiment demonstrate that the categorization rate of passive learning is 84.4 percent and that it is resistant to common cell phone postures and poses.