Abstract
Detecting left-behind passenger is crucial for preventing fatalities caused by high temperatures in enclosed car cabins. Due to non-line-of-sight (NLoS) challenges in complex vehicle interiors, existing wireless sensing approaches rely on distributed MIMO acoustic devices to address blind spots. However, these methods may still fail in challenging scenarios, such as when a passenger lies motionless on the rear seat facing backward or sleeps in the trunk. In this paper, we present PAX-Alert, a novel, full-coverage left-behind passenger detection system that operates using only a single pair of commodity speaker and microphone. By leveraging rich acoustic multipath reflections within the enclosed cabin, PAX-Alert is largely independent of the location and orientation of both the transceiver (speaker-microphone pair) and the passenger. It effectively detects both active passenger (awake with body movements) and static passenger (asleep or motionless) by capturing macro-body motions and subtle chest movement during respiration, which is main modulation source of multipath signals for static left-behind passenger. Extensive experiments demonstrate that PAX-Alert achieves high detection accuracy under standard NLoS conditions and remains effective even in extremely challenging NLoS scenarios. We believe PAX-Alert offers a flexible and robust solution for overcoming NLoS limitations in left-behind passenger detection.
| Original language | English |
|---|---|
| Journal | IEEE Transactions on Mobile Computing |
| DOIs | |
| Publication status | Accepted/In press - 1 Jan 2026 |
Keywords
- Acoustic sensing
- Left-behind passenger
- System multipath reflections
Fingerprint
Dive into the research topics of 'PAX-Alert: Full-Coverage Detection of Left-Behind Passenger in Car Cabin Using Single Speaker-Mic Pair'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver