Millimeter-Wave Radar-Based Bathroom Fall Detection using Time-Frequency Features

Falls in the elderly are a major public health problem. In this paper, an effective fall detection method based on millimeter wave radar system is proposed. Using the time-frequency characteristics of radar echoes, the signal-to-noise ratio is significantly improved by least squares adaptive filtering to eliminate interference, and classification is performed by combining short-time Fourier transform (STFT) and machine learning algorithms KNN (Nearest Neighbor), Support Vector Machine (SVM), Multilayer Perceptron (MLP) and Decision Tree (DT). By collecting data from two idle bathrooms, data from four actions were collected and the effectiveness of the machine learning SVM, KNN and MLP algorithms were verified by the measured data, in which the classification performance of SVM was better than the other three algorithms.