Gesture Recognition for Human-computer Interaction Based on CNN Model
Jing Cheng, Zhulin Li · 2021
Gesture recognition plays an important role in analyzing human-computer interaction. The existing gesture recognition methods mainly include hand postural morphological analysis and sensor data modeling. The former method can effectively identify hand gestures with large morphological differences in human-computer interaction, the latter use sensor devices to collect hand data for modeling. This paper proposes a gesture recognition method based on CNN visual model in human-computer interaction. The method does not require sensor to collect data, and can identify similar gestures in interaction process. The CNN model is constructed based on the VGG16 model, which extracts the system feedback sequence frames for each gesture action, constructs the frame sequence feature vector through the convolution layer of the CNN model, and finally predicts the probabilities of various types of gesture actions through the feature vector.