基于PCA降维的分层超限学习机手势识别方法

针对原始超限学习机在手势识别应用中欠缺良好的泛化性能和鲁棒性等问题,运用主成分分析(PCA)算法降低手势数据维数简化数据结构,并引入以超限学习机为基础,根据多层感知器理论拓展的分层超限学习机作为分类器应用于手势识别.PCA算法提取手势图像的主要特征,通过分层超限学习机的稀疏自动编码和分层训练,获得原始输入的多层稀疏表达,使自动编码后的输出近似原始输入,最大限度地减少重构误差,提高特征分类的精度.实验表明,与原始的超限学习机相比,具有更好的泛化性能,更快的识别速率以及更高的识别精度,提高了整体的学习性能.

Hierarchical extreme learning machine gesture recognition method based on PCA dimension reduction

The original extreme learning machine in application of gesture recognition,lacking of good generalization performance and robustness,it uses principal component analysis (PCA) algorithms to reduce gesture data dimension and simplify the data structure,and introduces hierarchical extreme learning machine as classifier applied in gesture recognition which is based on the extreme learning machine and according to expand the theory of multilayer perceptron.PCA algorithm extracts the main features of the gesture image,through a hierarchical extreme learning machine sparse automatic coding and hierarchical training,obtaines the original input multi-layer sparse expression to make the automatic encoding of the output approximate to the original input,maximize reduce reconstruction error and improve precision of classification features.Simulation results show that it has better generalization performance,faster learning speed and higher learning accuracy compared with the original machine,and improves the overall learning performance.