Temporal Learning using Echo State Network for Human Activity Recognition

Several works have been applied non-temporal classification techniques in the Human Activity Recognition area. Instead of that, we present an approach for modelling the human activities using a temporal learning tool. Here, the activities are considered as time-dependent events, and we use a temporal learning method for their classification. We employ a well-known learning tool named Echo State Network (ESN). An ESN is a specific type of Recurrent Neural Networks, which has proven well performances for solving benchmark problems with sequential and time-series data. Another advantage is that the method is very robust and fast during the learning algorithm. Therefore, it is a good tool for being applied in real time contexts. We apply the proposed approach for analyzing a well-know benchmark dataset, and we obtain promising results.