Proposing Gesture Recognition Algorithm Using Two-Stream Convolutional Network and LSTM

Identifying gestures and actions of human through video will help provide important and valuable information for interaction with surrounding devices as well as health care and entertainment activities. In recent years, many machine learning models have been proposed to identify human actions. In the paper, we propose a system to identify gestures of action of videos based on a two stream-convolution network (ConvNet) model. Besides, we also propose the method for recognizing human hand gestures in accordance with the purpose of the indoor environment. Our main contribution is to improve the two-stream ConvNet model based on [1]. The simulation results show that the proposal model has improved 20 percent of the processing speed and resources comparing with the existing models (VGG16, InceptionV3, Mobilenet V1, and Densenet201).