基于Wi-Fi CSI和胶囊网络的纺织纤维识别方法  

Textile fiber identification method based on Wi-Fi CSI and capsule network

作　　者：张慧卉 谷林 ZHANG Huihui;GU Lin(School of Computer Science,Xi’an Polytechnic University,Xi’an 710600,China)

机构地区：[1]西安工程大学计算机科学学院,陕西西安710600

出　　处：《传感器与微系统》2025年第3期107-110,116,共5页Transducer and Microsystem Technologies

摘　　要：针对传统纺织纤维识别方法中存在识别周期长、技术障碍高、检测仪器昂贵、且对专业人员依赖性强等问题,提出了一种基于Wi-Fi信道状态信息(CSI)的纺织纤维识别方法。首先,采集Wi-Fi信号的CSI并进行去噪处理;然后,提取小波包分解的时频特征,采用主成分分析(PCA)进行数据降维;最后,通过基于多头自注意力(MHSA)机制的胶囊网络(CapsNet)对输入特征矩阵的时空特征进行有偏向的提取,输出样本所属类别的概率。实验结果表明:该方法可以有效识别纺织纤维的种类,在室内独立环境下,平均识别率达到93.8%,证明了该方法的有效性和通用性,与现有的纺织纤维识别方法相比,具有更大的技术优势和更加广阔的现实应用前景。Aiming at the problems of long identification cycle,high technical barriers,expensive detection instruments,and strong dependence on professionals exist in traditional textile fiber identification methods,a textile fiber identification method based on Wi-Fi channel state information(CSI)is proposed.Firstly,the CSI of the Wi-Fi signal is collected and denoised.Then,the time-frequency features of wavelet packet decomposition are extracted,and principal component analysis(PCA)is used for data dimensionality reduction.Finally,the spatio-temporal features of the input feature matrix are extracted by capsule network(CapsNet)based on the mechanism of multi-head self-attention(MHSA)with a bias,and output probability of the category to which the sample belongs.Experimental shows that the method can effectively identify textile fiber categories,and the average recognition rate reaches 93.8%in an indoor independent environment,which proves the effectiveness and versatility of the method,and it has greater technical advantages and broader realistic application prospects compared with existing textile fiber identification methods.

关 键 词：纺织纤维识别 信道状态信息 胶囊网络 小波包分解 多头自注意力机制 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]