Paper Title
Convolutional Neural Network Based Human Activity Recognition
Abstract
The conventional methods of human activity recognition have achieved a tremendous growth. But the success of
these methods is heavily depended over the quality of manual feature extraction and hence these models suffer with the
problem of less generalizability. The widespread popularity and success of deep neural network attracts more research in
pervasive and ubiquitous computing around wearable sensor environment. In this paper a one-dimensional convolutional
neural network has been proposed for human activity recognition. Model was trained and tested on WISDM dataset which is
publicly available. The accuracy of 98.64% is achieved with this proposed architecture.
Keywords - CNN, Batch-normalization, human activity recognition, WISDM, Neural networks, wearable sensors