In this study, we compare Support Vector Machine (SVM), Convolutional Neural Network (CNN), and Long ShortTerm Memory (LSTM) algorithms as commonly used machine learning algorithms based on FMCW Radar data for Human Activity Recognition (HAR). The comparison is conducted by evaluating the models on test data and considering the fitting time and the number of parameters required by each model to achieve the desired results, to find the most efficient model that provides the best results. We discovered that the LSTM 01 model with one layer of 16 unit-LSTM produces the best result based on the scoring of several tested models. The model demonstrated an ability to achieve accuracy up to 86% on the test data with a relatively small number of parameters, i.e., 294,725. Key Words: Radar, computation, FMCW, SVM, CNN, LSTM.
Copyrights © 2023