Human Activity Recognition Using LSTM/BiLSTM

Abstract

The tracking and understanding the behavior of human beings is an important issue in a number of industrial applications. At the very best level, the system capable of resolving this problem has to be able to recognize the human behavior and understand the motive from that observation alone. This is a difficult task, as people can have different ways of performing certain tasks, so differentiating them properly can be challenging. In this work, a method for the design of a Deep Learning based intelligent human activity monitoring system is proposed, that can detect and track suspicious activity in any surveillance environment. This method makes use of features extracted using modern day image classification algorithms and passing them to sequence based neural network. Being able to perform classification of human activities on a live feed can be helpful in health sectors as well as surveillance systems and prevent disasters.