Classification Accuracy Improvement of Traffic Monitoring MW Radar by Velocity Compensation

In this study, target classification using machine learning for traffic monitoring systems with millimeter-wave radar has been studied. Time-frequency analysis by STFT is a strong tool for radar data, thereby we can see difference of periodicity in Doppler frequency. However, it is difficult to distinguish targets having weak periodic motion such as bicycles and cars. In traffic monitoring systems, observation time for target classification is also an issue. Since the speed of cars is high, target classification in short observation time is required. In this paper, we propose a mean-velocity compensation to spectrograms in classification by CNN. Effectiveness of velocity compensation in limited observation time is verified by using experimental dataset.