摘要 : This letter explores a critical facet of sensor technology: examining the feasibility of harnessing the preexisting WiFi infrastructure to monitor and assess driver behavior. We introduce WiFi improved safe driver operation monito... 展开 This letter explores a critical facet of sensor technology: examining the feasibility of harnessing the preexisting WiFi infrastructure to monitor and assess driver behavior. We introduce WiFi improved safe driver operation monitoring (WISDOM), a novel device-free action recognition system that utilizes commercial off-the-shelf (COTS) WiFi technology to enhance the safe operation of autonomous driving systems. WISDOM has two main phases: Phase I involves capturing channel state information (CSI) using a specialized monitoring system, and Phase II employs various data processing techniques, including phase offset elimination, outlier removal, and principal component analysis (PCA). In addition, a method called driving car waveform capture (DCWC) extracts relevant data from the CSI waveform. The system then employs long short-term memory (LSTM) networks and attention mechanisms to learn temporal action features, and random forest (RF) for classification. Experimental results demonstrate WISDOM's robustness, achieving impressive accuracy rates of 97.15% for driver identity recognition and 96.57% for action recognition. In a campus driving scenario, it outperforms recent methods with a 95.83% accuracy. 收起