Human Activity Recognition based on WaveNet

This paper makes a comprehensive study on human activity recognition based on CNN and RNN and proposes an alternative way to solve HAR based on WaveNet. The model is full probability autoregressive and can be used as a discriminant model. Therefore, this paper attempts to use WaveNet to solve HAR. By comparing the results of CNN, RNN and WaveNet models on the same data set, the experiment proves that WaveNet has strong superiority in some cases when dealing with HAR related problems, and the future development prospect is very optimistic.