A Fall Detection Method Based on K-Nearest Neighbor Algorithm with MIMO Millimeter-Wave Radar
Congzhang Ding, Zizhou Ding, Lingyu Wang, Yong Jia
Abstract
As a frequent accident in daily life, falls are the main factor that leads to physical and mental injury and even endangers life. Due to the independence and privacy of the home environment, the elderly living alone has no access to being helped after falls, which seriously threatens their lives. A fall detection algorithm based on MIMO millimeter-wave (mmWave) radar is proposed in this paper. The three manually extracted features of velocity, acceleration, and DOA change rate are utilized for the input of the K-nearest neighbor (KNN) algorithm to detect human falls, and the performance of the KNN algorithm using different parameters is verified. The experimental results show that the accuracy of human fall detection reaches 91.25% when the K value of the KNN algorithm is 5. Compared with the deep learning algorithm, the KNN algorithm effectively reduces the complexity.