Recognition of human actions based on deep convolutional neural network using postures and depth maps

Computer Engineering

The Human Action Recognition is one of the moat has become a significant space in the PC vision and furthermore has become an essential need for different PC applications that utilize individuals conduct.We can see the use of HAR in various fields of our everyday lives among which many come under surveillance (For traffic surveillance, hospital or bank activities,etc).As we have already seen or come through different methodologies already in existence for recognizing different human activities depending on postures.But there are very few that consider live input for action recognition.These are frameworks which are competent to perceiving the unpredictable human activity designs with a contribution from advanced camera and sensors. This implementation paper gives a useful and precise computation for various and distinct activity recognition. This paper presents a completely amazing part extraction and exhaustive gathering methodologies of the human exercises and getting ready measure. It essentially focuses to arrange the human movement plans with picture recuperation from the video input.The main difference between our system and already prevailing system is that out system is implemented on Deep - Convolutional neural network(D-CNN) by achieving an accuracy of 98.97% through image input,97.43 through video input and 79.07% through live input.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.