An Investigation of Material Classification Using 60 GHz mmWave FMCW Radar and Feature Selection-Based Machine Learning

This paper investigates the classification of various materials using 60 GHz mmWave Frequency-Modulated Continuous Wave (FMCW) radar and machine learning techniques. A feature extraction framework is developed, considering time-domain, frequency-domain, and phase-based features, followed by an evaluation of feature importance. The study explores the impact of feature selection on classification performance, utilizing random forest (RF), support vector machine (SVM), and k-Nearest Neighbors (kNN) classifiers. The results show that feature selection significantly improves classification efficiency, where a reduced feature set achieves comparable accuracy to full-feature models while reducing computational complexity. Performance comparisons using accuracy, precision, recall, F1-score, and ROC curves demonstrate the superiority of RF classifiers for material classification.