负重环境下基于足底压力和CNN-LSTM网络的性别识别方法研究  

Research on Gender Recognition Methods Based on Plantar Pressure and CNN-LSTM Network in Weight-bearing Environments

作　　者：姚井睿 杜明坤 王茜仪 YAO Jing-Rui;DU Ming-Kun;WANG Xi-Yi(Jiangsu Police Officer College,Nanjing Jiangsu210031,China)

机构地区：[1]江苏警官学院,江苏南京210031

出　　处：《机电产品开发与创新》2023年第6期13-15,共3页Development & Innovation of Machinery & Electrical Products

基　　金：江苏省公安厅科学研究计划项目(2019KX014);基于深度学习的足底压力自动分析方法研究(XJ202210329012)。

摘　　要：目前卷积神经网络(Convolutional Neural Networks,CNN)已被很广泛地用于步态识别领域,但CNN在分类时仅考虑单张静态步态图像,缺少对连续特征的关注,影响识别系统的准确率。因此,本文提出附加长短期记忆网络(Long Short-Term Memory networks,LSTM)来获取足底压力的连续特征。针对传统人工鉴定依赖主观经验的不足,及现有自动识别模式偏差较大、不能完全克服负重干扰的问题,本文采集不同负重环境下的足底压力数据,依据足底压力特征与男女性别差异,通过CNN-LSTM网络实现对足底压力进行个体性别的自动分类,算法识别率达到98.31%。Currently,Convolutional Neural Networks(CNN)have been widely used in the field of gait recognition.However,CNN only considers a single static gait image when classifying,lacking attention to continuous features,which affects the accuracy of the recognition system.Therefore,this article proposes adding Long Short Term Memory Networks(LSTM)to obtain continuous features of plantar pressure.In response to the shortcomings of traditional manual identification relying on subjective experience,as well as the large deviation of existing automatic recognition modes that cannot fully overcome weight interference,this article collects plantar pressure data under different weight bearing environments.Based on plantar pressure characteristics and gender differences,the CNN-LSTM network is used to achieve automatic gender classification of plantar pressure,with an algorithm recognition rate of 98.31%.

关 键 词：卷积神经网络 足底压力 步态特征 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]