Research on multi-user identity recognition based on Wi-Fi sensing
With the advancement of wireless sensing technology, research on Wi-Fi-based identity recognition has garnered significant attention in fields such as human-computer interaction and home security.While identity recognition based on Wi-Fi signals has achieved initial success, it is currently primaril...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | zho |
| Published: |
China InfoCom Media Group
2024-03-01
|
| Series: | 物联网学报 |
| Subjects: | |
| Online Access: | http://www.wlwxb.com.cn/zh/article/doi/10.11959/j.issn.2096-3750.2024.00381/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | With the advancement of wireless sensing technology, research on Wi-Fi-based identity recognition has garnered significant attention in fields such as human-computer interaction and home security.While identity recognition based on Wi-Fi signals has achieved initial success, it is currently primarily suitable for scenarios involving individual user behavior.Identity recognition for multiple users in concurrent behavior scenarios still faces a series of challenges, including issues related to mutual interference between users and poor model robustness.Therefore, a Wiblack system for recognizing multiple user identities in a concurrent distribution behavior scenario was proposed.The core idea was to train a multi-branch deep neural network (Wiblack-Net) to extract unique features for each individual user.Firstly, the common features among multiple users were extracted using the backbone network.Then, a binary classifier was assigned to each user to determine the presence of the target user within a given group, thereby achieving identity recognition for multiple users based on concurrent behavior.In addition, experiments comparing Wiblack with several independent binary classification models and a single multiclassification model were conducted to analyze operational efficiency.System performance experimental results demonstrate that when simultaneously identifying the identities of three users, Wibalck achieves an average accuracy of 92.97%, an average precision of 93.71%, an average recall of 93.24%, and an average F1 score of 92.43%. |
|---|---|
| ISSN: | 2096-3750 |