< Projeler Sayfası

Proje Adı
μ-Doppler Radar Kullanarak İnsan ve Hayvan Hareketlerinin Tespiti ve mikro-Doppler İmzalarının Karşılaştırmalı İncelenerek Sınıflandırılması

Şirket Adı
Meteksan Savunma

Şirket Mentoru
Cesur Karabacak

Akademik Mentor
Prof. Enis Çetin

Asistan
Ersin Yar

Takım Üyeleri
Lizge Gül Uludağ, Mehmet Erhan Erköseoğlu, Elif Agah, Canberk Hallik, Gamze Onuker

Summary
The system provides the detection and classification of human movements using a radar. Walking, running, creeping and crawling are the main human movement types examined, where they are detected and classified according to signatures of radar signal returns. If transmitted signal with a certain frequency hits an object in motion, a shift can be observed on the radar receiver according to radar signal returns. This situation is named Doppler Effect. If the object has oscillating parts such as legs, arms and torso of humans, they will form difference frequency shifts on Doppler Effect. This is named micro-Doppler Effect. Each moving human has micro-Doppler signatures unique to himself and this situation provides classification of different human movements. Micro-Doppler signature detection and classification provides wide application opportunities for industrial areas from defense industry to medical industry, which was the motivation behind the initiation of this project. For the classification of these signals, Support Vector Machines and k-Nearest Neighbor algorithms are used. The outputs of Mel Filtered extracted data and envelope data are fed into these classification systems, which then provide the type of movement that is decided by the system. For the simulation data, we have reached above 95% accuracy with Mel Filtered data. The radar used in the project has a low Doppler pulse frequency, which is not sufficient for micro-Doppler signature evaluation. However, the methods we developed for the simulations, added with the new algorithms we developed to tackle real life problems, have provided promising results in classification.