一种VR游戏手势运动识别装置

Gesture recognition device of VR game

设计一种VR游戏手势动作识别装置,基于惯性传感器芯片识别从持握初始点向上挥动、向下挥动、向左挥动、向右挥动四种手势。该装置的硬件部分使用STM32单片机为核心微处理器,搭载MPU6050空间运动传感器芯片以采集手势运动数据,通过WiFi模块上传手势特征数据到游戏服务器,游戏服务器运行分类模型进行手势分类识别。软件部分着重改进手势数据特征提取方法,开创性地提出连续加速度向量幅值作为数据判断依据,并创新使用拟合算法在不同的手势数据中按照时间序列提取出相同长度的特征数据矩阵。在多种机器学习算法中选择线性判别分析算法作为手势分类识别算法,采集340个手势特征数据作为算法训练集和109个手势特征数据作为算法测试集,测试结果获得了98.17%的识别准确率。实验结果表明,所设计装置具有硬件成本低、使用灵活、功耗小、准确率高等特点,运用在VR游戏中可以改善人机交互过程。

A gesture recognition device for VR(virtual reality)games is designed.Based on the inertial sensor chip,It can recognize four gestures:waving upward,waving downward,waving towards the left,and waving towards the right from the initial point of holding.In the hardware part of this device,microcontroller STM32 is used as the core microprocessor,space motion sensor chip MPU6050 is equipped to collect gesture motion data,the gesture feature data is upload to the game server through the WIFI module,and the game server runs the classification model to perform gesture classification recognition.The software part of this device focuses on improving the feature extraction method of gesture data.The continuous acceleration vector amplitude is initiatively proposed as the basis for data judgment,and a fitting algorithm is innovatively used to extract feature data matrixes with same length in different gesture data according to time series.In a variety of machine learning algorithms,the linear discriminant analysis algorithm was selected as the gesture classification and recognition algorithm.340 gesture feature data were collected as the algorithm training set and 109 gesture feature data were used as the algorithm test set.The test result of 98.17%recognition accuracy was obtained.The experiment result shows that the designed device has the characteristics of low hardware cost,flexible use,low power consumption,high accuracy,etc.It can be used in VR games to improve the humancomputer interaction process.