Classification and Regression of Ultra Wide Band Signals

The Ultra-Wide Band (UWB) signals recently have attracted increasing attention in the area of material identification due to their potential of providing very high data rates at relatively short ranges and their capability of being obtained nondestructively and contactless. The Support Vector Machines (SVM) offers one of the most robust and accurate classification capability among the well-known such algorithms. In this paper, the SVM is applied in classifying different sets of high dimensional UWB signals that are collected from various liquid materials. The Support Vector Regression (SVR) and Artificial Neural Network (ANN) are also tested to predict the concentration of liquid using UWB. The results demonstrate that the SVM is an effective tool for differentiating materials by UWB, and SVR and ANN are acceptable in predicting UWB signals.