基于IFOA-ELM的车载手势识别方法研究

Research on Vehicle Gesture Recognition Method Based on IFOA-ELM

车载手势识别作为一种人机交互方式是提高道路行驶安全性的有效途径.针对传统车载手势识别方法研究中识别准确率和效率较低以及性能不稳定的问题,提出一种改进的果蝇优化算法(IFOA)优化极限学习机(ELM)参数的车载手势识别新方法.首先利用IFOA优化ELM的初始权重w和偏置b;接着采用最佳初始权重和偏置来训练ELM;最后利用IFOA-ELM对提取的车载手势特征向量进行手势类型识别;实验结果表明,与SVM、动态贝叶斯网络、传统ELM、FOA-ELM等分类学习算法相比,方法在高效稳定的前提下取得更高的识别准确率,满足对准确性和实时性要求较高的车载环境中的手势识别.

Vehicle gesture recognition as a way of human-computer interaction is an effective way to improve the safety of road driving.However,existing vehicle gesture recognition algorithms are less accurate and inefficient and performance is unstable.We propose an novel vehicle gesture recognition method based on IFOA-ELM.Firstly,Improved Fruit Fly Optimization Algorithm(IFOA) is utilized to optimize the input weight w and bias b of Extreme Learning Machine(ELM).Then,the ELM is trained by the best input weight and bias.Finally,the IFOA-ELM is used to identify the gesture type of the extracted vehicle gesture feature vector.The experimental results show that compared with the classification learning algorithms such as SVM,dynamic bayesian network,traditional ELM and FOAELM,our method achieves higher recognition accuracy under the premise of high efficiency and stability,and satisfies the gesture recognition in vehicle environment with high accuracy and real-time requirement.