Multiple Target Vital Sign Detection Using Ultra-Wideband Radar

Detection of living beings through wall or under the rubble/debris is difficult in Radar due to low signal to noise ratio (SNR). To improve SNR and extract vital signs from Radar echo signal advanced signal processing techniques are required. The existing methods focus on detection of vital sign for single target only. This paper proposes an improvised method to detect multiple vital signs by determining the number of targets and their breathing frequency behind the wall using fast Fourier transform (FFT). Additionally an alternative technique using deep learning has been proposed for faster pre-processing and classification of number of targets. The proposed convolutional neural network (CNN) model is compared with other machine learning classifiers. It is found that if we used CNN as a feature extractor layer and Random Forest (RF) classifier it works the best with 94.38% accuracy (for two targets) and 73.33% accuracy (for three targets) without any pre-processing on Radar data. This method reduces the detection time needed as compared to the conventional algorithms available in literature