The Ultra Wide Band Signals (UWB) has recently attracted increasing attention in the area of material identification because of its potential to provide very high data rates at relatively short ranges and it can be obtained nondestructively and contactless. As a result, the reflected UWB signals offer one potential contactless material identification or classification tool. In this research, the UWB signals collected in a series of liquid material classification tests are studied. Then the spectral clustering algorithms of Ng-Jordan-Weiss (NJW) and LIHI (A new algorithm discovered by Lihi Zelnik-Manor) are applied to group the UWB signals into some desired number of classes. The outputs show that Spectral Clustering algorithm of NJW and LIHI have perfect performance in distinguishing UWB signals. The Support Vector Machine (SVM) is a machine learning method proposed by Vapnik which is based on statistical learning theory and it offers one of the most robust and accurate classification and regression capability among all well-known such algorithms. In this thesis, SVM is applied in classifying different sets of UWB signals as well. The Support Vector Regression (SVR) and Artificial Neural Network (ANN) are also tested. As a result, the outstanding performance of classification proves that SVM is an effective tool for differentiating materials. SVR and ANN are reasonable in predicting UWB signals. The data preprocessing of Principle Component Analysis (PCA), demodulation and range selection are also equivalent important in this research and it will be explored at the same time