Ultra-wideband Impulse Radar Classification of Materials using a Convolutional Neural Network

We present an application of ultra-wideband impulse (UWB) radar for material classification. A single input-single output (SISO) which has a center frequency of 7.25-GHz was used. Eight different materials and objects, which are book, note, copper, aluminum, plastic, CD, and cup were used to evaluate the capability of the SISO radar. A deep convolutional neural network model was developed and trained directly on the raw UWB radar waveform signal. The results show that the SISO UWB radar could classify well with an average accuracy over 90%, higher than some other machine learning models such as support vector machine, Gaussian Naïve Bayes. The method could be extended to apply in real-life applications such as object recognition in manufacturing quality control, hidden object detection, and material properties search.