基于PSO改进的BP神经网络数据手套手势识别

Gesture recognition of data glove based on PSO-Improved BP neural network

针对5DT数据手套手势识别这一问题,提出BP神经网络和PSO算法相结合的识别方法。首先利用特征提取和归一化方法建立通用手势模板,并基于此模板采用BP神经网络进行训练学习,同时通过PSO算法修正BP神经网络的权值和阈值,将训练完毕的神经网络用于实际操作过程中的手势识别。该方法既保留了BP算法结构简单、易于实现的优点,同时避免了不同操作者复杂的标定过程。仿真和实验结果表明,所提出的控制方法有效的缩短了学习时间,并且提高了识别过程的实时性和精确性。

For the 5DT data glove gesture recognition,the recognition method that combines the BP neural network with PSO （Particle Swarm Optimization） algorithm was proposed.Feature extraction and normalization method was adopted to establish the common gesture template.The BP neural network was trained to learn the established gesture template.Meanwhile the weights and thresholds of the BP neural network were modified by PSO algorithm.Then the trained neural network was used to recognize gestures in actual manipulation process.This method avoids the complex calibration of different operators.The advantages of BP algorithm,including simple structure,easy realization are retained,and the real-time performance and accuracy in the identification process are improved with the introduction of the PSO algorithm.The simulation and experimental results show that the proposed control method shortens the learning time efficiently and improves the real-time performance and accuracy of recognition process.