WiCAR wifi-based in-car activity recognition with multi-adversarial domain adaptation

In-car human activity recognition is playing a critical role in detecting distracted driving and improving human-car interaction. Among multiple sensing technologies, WiFi-based in-car activity recognition exhibits unique advantages since it does not rely on visible light, avoids privacy leaks and i...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Proceedings of the International Symposium on Quality of Service S. 1 - 10
Hauptverfasser: Wang, Fangxin, Liu, Jiangchuan, Gong, Wei
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: New York, NY, USA ACM 24.06.2019
Schriftenreihe:ACM Other Conferences
Schlagworte:
Activity recognition
Adaptation models
Automobiles
Deep Learning
Domain Adversarial Network
Feature extraction
Hardware > Robustness
Human-centered computing > Ubiquitous and mobile computing > Ubiquitous and mobile computing systems and tools
In-Car Human Activity Recognition
Measurement
Networks > Network components > Wireless access points, base stations and infrastructure
Networks > Network performance evaluation > Network performance modeling
WiFi Signal Processing
Wireless fidelity
in-car human activity recognition
deep learning
domain adversarial network
wifi signal processing
ISBN:9781450367783, 145036778X
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:In-car human activity recognition is playing a critical role in detecting distracted driving and improving human-car interaction. Among multiple sensing technologies, WiFi-based in-car activity recognition exhibits unique advantages since it does not rely on visible light, avoids privacy leaks and is cost-efficient with integrated WiFi signals in cars. Existing WiFi-based recognition systems mostly focus on the relatively stable indoor space, which only yield reasonably good performance in limited situations. Based on our field studies, the in-car activity recognition, however, is much more complicated suffering from more impact factors. First, the external moving objects and the surrounding WiFi signals can cause various disturbances to the in-car activity sensing. Second, considering the compact in-car space, different car models can also lead to different multipath distortions. Moreover, different people can also perform activities in different shapes. Such extraneous information related to specific driving conditions, car models and human subjects is implicitly contained for training and prediction, inevitably leading to poor recognition performance for new environment and people. In this paper, we consider the impact of different domains including driving conditions, car models and human subjects on the in-car activity recognition with field measurements and experiments. We present WiCAR, a WiFi-based in-car activity recognition framework that is able to remove domain-specific information in the received signals while retaining the activity related information to the maximum extent. A deep learning architecture integrated with domain adversarial training is applied to domain independent activity recognition. Specifically, we leverage multi-adversarial domain adaptation to avoid the discriminative structures mixing up for different domains. We have implemented WiCAR with commercial-off-the-shelf WiFi devices. Our extensive evaluations show that WiCAR can achieve in-car recognition accuracy of around 95% in untrained domains, where it is only 53% for solutions without domain adversarial network and 83% for the state-of-the-art domain adversarial solution.
ISBN:9781450367783
145036778X
DOI:10.1145/3326285.3329054