human activity recognition using CNN and long term short term memory

Human activity recognition aims to work out the activities performed by someone in a picture or video. Examples of actions are running, sitting, sleeping, and standing. Complex movement patterns harmful occurrences like falling may be part these activities. The suggested ConvLSTM network can created successively combining fully connected layers, long immediate memory (LSTM) networks, convolutional neural networks (CNN). acquisition system will pre calculate skeleton coordinates using human detection pose estimation from image/video sequence. model builds new controlled features raw their distinctive geometric kinematic properties. Raw utilized generate properties supported relative joint position values, differences, angular velocities. By utilizing multi-player trained CNN-LSTM combination, novel spatiotemporal directed obtained. classification head with completely layers is then utilized. was tested KinectHAR dataset, which consists 130,000 samples 81 attribute variables compiled Kinect (v2) sensor Experimental data used compare performance independent CNN LSTM networks.