基于步态特征的性别识别方法研究  

Research on Gender Classification Based on Gait Features

作　　者：李忠浩 罗文田[1] 陈乾 覃华懋 Li Zhonghao;Luo Wentian;Chen Qian;Qin Huamao(Civil Aviation Flight University of China,Guanghan 618307 Sichuan China)

机构地区：[1]中国民用航空飞行学院,四川广汉618307

出　　处：《中国民航飞行学院学报》2025年第1期59-64,共6页Journal of Civil Aviation Flight University of China

摘　　要：本文基于步态特征设计了一套人物性别识别模型。以能量图的形式作为该模型的输入,模型采用串联式的网络结构并利用Softmax函数进行最终的人物性别识别。然后将CASIA-B数据库中的步态轮廓图裁剪合成能量图输入模型测试性别识别得到91%以上的整体识别准确率。最后设计实验对比本文模型和VGG、ResNet识别模型,实验结果表明该模型针对人物性别识别有着更好的泛化性能和效率。This paper designs a set of gender classification model based on gait features.The input of the model is in the form of an energy graph.The model adopts a series-linked network structure and the Softmax function for the final character gender classification.Then,the gait profile in the CASIA-B database is clipped to synthesize an energy profile and input into the model for gender classification testing.The overall classification accuracy is more than 91%.Finally,an experiment is designed to compare the proposed model with the VGG and ResNet classification models.The experimental re-sults show that the model has better generalization performance and efficiency for character gender classification.

关 键 词：神经网络 步态特征 能量图 性别识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]