Radar-Based Fatigue Driving Action Recognition Using the Combination of Micro-Doppler Time-Frequency and Energy Features

The accurate recognition of driver actions has significant practical significance for ensuring driving safety and optimizing driving experience. Radar sensors have shown great potential for application in monitoring driver behavior, owing to their advantages of noncontact, privacy protection, high precision, and high weather resistance. For high-accuracy fatigue driving action recognition, this study proposes a radar-based fatigue driving action recognition method using the combination of micro-Doppler time-frequency and energy-distribution features, a dual-stream fusion CSPR Net model. We first use a radar sensor to collect motion echo data, followed by noise suppression using preprocessing algorithms, such as fast Fourier transform and order statistic-constant false alarm rate algorithms, to generate micro-Doppler time-frequency and energy-distribution maps of human movements. Subsequently, we extract multidimensional feature information, such as velocity, time frequency, and energy of human movements, through CSP and transmission modules. Finally, we perform weighted fusion on the multidimensional feature information to achieve motion recognition. The experimental results show that the proposed algorithm achieves an average classification accuracy of 99.5% for seven driver actions in known scenarios and 95% in unknown generalized scenarios, demonstrating good robustness.