A Research of Report Analysis of Gestures in VR Games

This article, through literature review and theoretical analysis, points out the advantages and limitations of gesture interaction in terms of operational naturalness and scene adaptability, and provides reasonable improvement ideas and the major trends in the future direction of VR. It focuses on analyzing five typical gesture recognition technologies, such as the data glove technology based on images, which uses the AR toolkit on fingertips and inverse kinematics technology to infer the movement trajectory of the user's hand. RGB image 3D hand pose recognition technology, which analyzes a single RGB image to infer 2D and 3D images and guess the player's gesture situation. Sparse multi-channel surface electromyography (sEMG) recognition uses a 4-stream model architecture, with a Butterworth filter for noise reduction during preprocessing, and μ-law transformation to amplify weak signals, along with channel attention to amplify features, maintaining high accuracy on the DB1 and DB9 datasets. The DTW-RCE neural network recognition technology for IMU sensors processes time-series data collected by inertial measurement units (IMUs). In the data processing flow, a 9-axis IMU sensor is used as the data acquisition end, synchronously obtaining three-axis acceleration and three-axis angular velocity data at a sampling frequency of 100Hz. The IMU and acoustic sensor multimodal recognition technology enhances robustness by integrating wrist acoustic signals, recording limb movements with three-axis accelerometers and gyroscopes. At the same time, it summarizes the existing challenges in real-time performance, environmental adaptability, and generalization ability of the technology, and proposes optimization solutions.