基于深度学习与步态分析的身份识别算法

Identity recognition algorithm based on deep learning and gait analysis

在传播性公共卫生疫情环境下,为了减少传染风险,医疗机构需要对医护人员和患者进行防护状态下的身份识别。文中基于深度学习算法提出了一种步态识别架构顺序残差卷积网络(SRCN),用来提取基于卷积主干的时空信息,从而实现对个体行走模式的学习。利用信息提取器(BIE)和多帧聚合器(MFA)两个子模块对图像时空信息进行提取,使用残差神经网络(ResNet)提取每幅图像的空间特征。MFA将整合并提取所有特征从而实现步态识别。在CASIA-B公开数据集进行的实验表明,文中所提出的方法在3种模态下准确率分别达到了95.2%、89.8%和73.9%,相比其他模型均有所提高。

In the environment of communicable public health epidemic,in order to reduce the risk of infection,medical institutions need to identify the medical staff and patients in the state of protection.In this paper,using deep learning algorithm,a gait recognition architecture,Sequential Residual Convolution Network(SRCN),is proposed to improve the extraction of temporal and spatial information based on convolution backbone,so as to realize the learning of individual walking pattern.Bit Information Extractor(BIE)and Multi Frame Aggregator(MFA)are used to extract the spatiotemporal information of the image,Residual neural Network(ResNet)is used to extract the spatial features of each image.MFA will integrate and extract all features to achieve gait recognition.Experiments on CASIA-B open data set show that the accuracy of the proposed method in three modes is 95.2%,89.8%and 73.9%respectively,which is better than other models.