Objective: This study was conducted to identify key features for efficient motion recognition performance prior to actual radar measurements, as part of the development of a markerless motion analysis algorithm using radar. The aim of this study is to extract key radar features for developing an efficient markerless method by utilizing radar signals calculated from marker coordinates measured by a 3D motion analysis system, classifying individual movements, and analyzing various factors. Method: 2 male (age: 25.00 ± 0.00 years, height: 174.00 ± 8.49 cm, weight: 77.00 ± 15.56 kg) and 3 female (age: 25.00 ± 2.65 years, height: 164.83 ± 2.75 cm, weight: 56.00 ± 7.00 kg) participated in this study and the marker coordinate of each subject engaged in 5 different motions were obtained. Assuming the continuous wave radar, the signal was modeled using the marker coordinate and classification was conducted by using the feature obtained from the correlation of time-frequency images (TFIs); TFI of the motion was modeled by using the entire markers and the feature was obtained by correlating the TFI of each marker to the TFI of the motion. Results: In this study, a classification accuracy of 92.32% was achieved under a Signal-to-Noise Ratio (SNR) condition of 0 dB, with near-perfect classification results across various SNR values, demonstrating the effectiveness of the proposed method. Simulation results showed classification accuracy of = 98% with a signal duration of = 0.5 sec and a sampling interval of = 0.25 sec in the training database. This suggests that the signal duration is a more critical factor for classification performance than the sampling interval. Conclusion: Our simulation results demonstrate that the motion can be classified by using the efficient feature obtained by the correlation between TFI of the motion and that of each marker stored in the training database.
© Copyright 2024 Korean Journal of Sport Biomechanics. Korean Society of Sport Biomechanics. Alle Rechte vorbehalten.