摘要
基于WiFi信道状态信息(CSI)的步态识别作为一种新型的生物特征识别方案,从步态周期、步速、行走姿势、身体平衡、足底压力等方面的差异进行分析,从而有效识别行人身份。然而,目前大多数基于WiFi CSI的步态识别方案中主要针对人体的运动特征进行识别,鲜见对人体物理形态特征进行分析,这样就导致了特征提取不完整,系统识别率低,鲁棒性差等问题。针对上述情况,提出了一种结合步行时运动特征和相对静止时物理形态特征的多特征融合身份识别方法。该方法利用了一个双流卷积神经网络(CNN)分别提取粗粒度运动特征和细粒度物理形态特征,并将两部分特征融合,进行身份识别。在真实的场景中进行了实验,最高准确率达到97.1%,明显优于现有的识别方案,实验证明了该方法的有效性和通用性。
As a new type of biometric recognition solution, gait recognition based on WiFi channel state information(CSI)is analyzed from the differences in the gait cycle, gait speed, walking posture, body balance, and plantar pressure, so as to effectively identify pedestrians.However, most of the current gait recognition solutions based on WiFi CSI mainly focus on human motion features, but rarely analyze the physical morphological features of the human body, which leads to problems such as incomplete feature extraction, low recognition rate, and poor robustness of the system.To address the above situation, a multi-feature fusion identity recognition method is proposed that combines motion features while walking and physical morphological features while relatively stationary.The method utilizes a dual-flow convolutional neural network(CNN)to extract coarse-grained motion features and fine-grained physical morphological features respectively and fuses the two parts of features for identity recognition.Experiments are conducted in real scenarios and a maximum accuracy of 97.1 % is achieved, which is significantly prior to existing recognition schemes, demonstrating the effectiveness and generality of this method.
作者
吴文杰
朱耀麟
梁颖
WU Wenjie;ZHU Yaolin;LIANG Ying(School of Electronics and Information,Xi’an Polytechnic University,Xi’an 710600,China;School of Computer Science,Xi’an Polytechnic University,Xi’an 710600,China)
出处
《传感器与微系统》
CSCD
北大核心
2023年第3期144-147,151,共5页
Transducer and Microsystem Technologies
基金
陕西省教育厅重点科研计划资助项目(20JS052)
陕西省自然科学基础研究计划资助项目(2021JM—453)。
关键词
步态识别
WiFi信道状态信息
运动特征
物理形态特征
双流卷积神经网络
gait recognition
WiFi channel state information(CSI)
motion features
physical morphological features
dual-flow convolutional neural network(CNN)
