Research on Gesture Recognition Based on Neural Network and Virtual Reality Technology

With the rapid development of computer vision technology and virtual reality technology, gesture recognition has become an important way to realize human-computer interaction. In the face of increasingly complex application scenarios, background noise, self-occlusion, illumination changes and other issues will have a great impact on the accuracy of gesture recognition, which restricts the application effect of gesture recognition interactive technology. In view of this, this paper will integrate artificial neural network technology, computer vision technology and virtual reality technology, and propose a set of gesture recognition framework based on convolutional neural network, which provides a new idea for accurate and rapid gesture recognition in complex background. Practice has proved that the key points of the hand in the picture can be effectively identified by MediaPipe method, and the three-dimensional coordinates of the corresponding key points can be output. Then combined with the gesture classification model of YOLOv5 network, the input image can be predicted and the image category can be output, so that the purpose of gesture intelligent recognition can be achieved. The overall gesture recognition framework has good stability, can extract the robust features of images in complex background, improve the accuracy of gesture recognition, and effectively realize human-computer interaction in virtual environment.