Motion Classification Using CNN Based on Image Difference
Wafaa Shihab Ahmed, Abdul Amir Abdullah Karim · 2020
The classification of human actions has become an important topics in recent researches. Typically the function of recognition human action is converted to the function of classifying the image that represents the person’s motion. In this paper to classify the human motion the Convolution Neural Network (CNN) has been used to extract features by convolutional layers and in fully connected layer Softmax classifier is used to classify the motion. This method evaluate the differences between two sequences frames and this frame differences is used for training and testing in CNN. The propose system has been applied on three databases KTH, Ixmas and Weizmann. The results of experiments achieved accuracy 98.75% with KTH, 92.24% with Ixmas and 100% with Weizmann database.