Human motion recognition based on packet convolution neural network
Yupeng Ding, Hongjun Li, Zhengyu Li · 2017
In order to solve the confusion of input data, an algorithm of human action recognition based on packet convolution neural network is proposed. The two-layer wavelet combined with the mean square error method is used to group the samples, and then study the features in the case of guaranteeing the grouping error. The algorithm is tested on the video library and compared with the traditional convolution neural network algorithm. The experimental results show that the proposed algorithm has a significant improvement in the subjective and objective performance compared with the similar algorithm, and the success rate has been greatly improved.