基于图像的异常步态二次特征提取算法仿真

针对传统特征提取算法容易忽略对数据的降维处理,未能较好的提取出图像异常步态特征,导致提取准确率不高的问题,提出一种基于图像的异常步态二次特征提取算法。根据不同姿态下足底的压力变化数值,完成异常步态图像的一次特征提取;根据一次特征提取后正常步态定义的特征变量数据,构建目标个体行走轮廓的步态能量图,同时结合KPCA核方法,对一次特征提取后的步态轮廓数据进行降维处理,完成异常步态的二次特征提取。仿真结果表明,在正常步态下和异常步态下,所提算法都能够有效地提取出图像的异常步态,具有较高的特征提取准确率,表明所提算法具有较强的实用性。

Simulation of Abnormal Gait Quadratic Feature Extraction Algorithm Based on Image

Traditional feature extraction algorithms are easy to ignore the dimensionality reduction of data. It cannot extract the abnormal gait features of image well, resulting in the low accuracy of extraction. Therefore, an algorithm to extract abnormal gait quadratic feature was presented. According to the change of plantar pressure in different postures, the feature extraction of abnormal gait image was completed. According to the feature variable data defined by the normal gait after the first feature extraction, the gait energy image of the walking contour of targeted individual was constructed. Meanwhile, KPCA kernel method was used to reduce the dimension of gait contour data after the first feature extraction. Finally, the secondary feature extraction of abnormal gait was completed. Simulation results show that the proposed algorithm can extract the abnormal gait from the image effectively, which has high accuracy of feature extraction. Thus, the proposed algorithm has strong practicability.