Radar-Tabanlı Düşme Tespitinde Güç Spektral Yoğunluk Özellikleri için Ampirik Mod Ayrıştırması

In recent years, the increase in the number of older people and their tendency to live alone has made them more vulnerable to accidents. The most suffered situation in this regard is their falls. In this study, fall detection is carried out using radar. The proposed method classifies different falls and activities of daily living using radar-based measurements. The signals obtained by means of empirical mode decomposition (EMD) are separated into intrinsic mode functions (IMFs). The power spectral densities (PSD) of IMFs are calculated using the Welch method to provide features for classification. Thus, the effect of IMFs on classification is observed. In the study, conventional machine learning classes are employed, and the Support Vector Machine (SVM) (cubic) classifier detects the fall with 100% accuracy as a result of the PSDs calculated depending on the IMF 2-6 values. Furthermore, the classification results obtained based on other IMFs are almost error-free for some classifiers. Therefore, classification is also performed for seven different movements depending on IMFs. The SVM (cubic) algorithm performs above 90% in this case. The proposed method demonstrates that the effect of classical machine learning remains operative and efficacious.