Support Vector Machine-Based Instantaneous Presence Detection for Continuous Wave Radar Systems

Unpublished / Preprint

Publication Details

Author(s): Schellenberger S, Shi K, Steigleder T, Michler F, Lurz F, Weigel R, Kölpin A
Publication year: 2019
Language: English


Instantaneous detection of missing vital signs at inpatient beds enables fast intervention for cardiac arrests.
Using a 24 GHz bistatic radar, a fast presence detection based on a support vector machine (SVM) classifer is realized. Large body motions or even small distance deviations, such as movement of the chest induced by heartbeat or breathing, are distinguishable from the measured noise of an unoccupied bed. For classifcation two features are calculated based on windowed I and Q data. Performance is evaluated by varying window sizes from 0.2 ... 1.5 s for feature calculation and training of the SVM classifer. In the resting scenario an accuracy of 99.2% and F1-score of 99.1% with windows of 0.2 s is achieved.

FAU Authors / FAU Editors

Lurz, Fabian
Lehrstuhl für Technische Elektronik
Michler, Fabian
Lehrstuhl für Technische Elektronik
Schellenberger, Sven
Lehrstuhl für Technische Elektronik
Shi, Kilin
Lehrstuhl für Technische Elektronik
Weigel, Robert Prof. Dr.-Ing.
Lehrstuhl für Technische Elektronik

External institutions
Brandenburgische Technische Universität Cottbus-Senftenberg

Last updated on 2019-11-01 at 20:23