Study of radar signature extraction for effective gesture classification with machine learning

Modern radar technology has various types of applications and brings convenience to people from different perspectives. Radar gesture recognition can be one of them. With the help of machine learning, it can reach a reliable classification and recognition rate. This project is aimed to compare the influence of different radar spectrogram feature extraction methods on recognition accuracy, focusing on finding the suitable feature extraction method under different scenarios and with different machine learning algorithms. This report summarizes the knowledge of micro-Doppler radar, image processing, feature extraction method as well as training algorithms. As a result, principal component analysis (PCA) together with AlexNet produced the best accuracy up to 100% with certain hand gesture radar spectrogram dataset.Bachelor of Engineering (Electrical and Electronic Engineering