Upload
buikhue
View
213
Download
0
Embed Size (px)
Citation preview
Radar Signal Processingfor Fall Motion Detection
Pro-/Projektseminar or Bachelor/Master Thesis
Detecting falls of elderly persons remains a challenging problem. Here, radar provides a non-obstrusive,safe and reliable technology to detect fall motions. In the near future, radar signal processing for fallmotion detection will be a key technology for assisted living. Radar systems as small as the palm of yourhand will be installed in apartments to monitor the motions of the elderly person.
Each motion induces characteristic Doppler frequency shifts in the radar return signal. They form theso-called micro-Doppler signatures in the time-frequency domain. These micro-Doppler signatures canbe used to classify the motion, i.e. to identify whether a person is walking normally, limping or walkingwith a cane for example. However, motions such as fast sitting and falling can be mistaken for one an-other as both motions reveal similar Doppler signatures. Thus, the challenge remains to classify humanmotions correctly and, in particular, reduce the false alarm rate in fall motion detection.
Students participating in this project will learn about detection and estimation theory focusing on micro-Doppler analysis. In particular, the project is about radar-based human gait analysis and fall motiondetection. Students will learn about time-frequency analysis for signal processing. Using real mea-surement data, we consider the short-time Fourier transform or spectrogram to discriminate differentmicro-Doppler signatures. In order to classify the motion, features are extracted from the spectrogramand fed to a classifier. Taking the spectrogram as an image, also image processing techniques can beused for feature extraction and classification. Depending on the student’s knowledge and interest, anindividual topic for a student project can be discussed.
SIGNAL
PROCESSING
GROUP
If you are interested in the project please contact:
Ann-Kathrin Seifert, [email protected] S3|06 252
Image sources: iStock.com/SilviaJansen (left).