Principal Component Analysis for Feature Extraction from One Dimensional Signals

This proposal responds to the solicitation for methods to capture features from one dimensional sensor signals. We exploit a combination of Principal Component Analysis (PCA) and Learning Vector Quantization (LVQ) to capture the required features. This can be used to diagnose a wide variety of systems using sensors which are monitoring the various parameters of the system. We will also describe an innovative one dimensional test probe we have developed that is synergistic with the PCA and LVQ algorithms to yield a very powerful diagnostic tool. The goal of this work is to develop a toolbox of PCA and LVQ implementations optimized for equipment diagnostics. The toolbox will also include software optimized for use with our innovative one dimensional test probe signal concept, but would be equally useful for interpretation of any one dimensional signal. We will also show examples later of the PCA and LVQ algorithms used for a wide range of other applications where the input data set has thousands of degrees of freedom so our toolbox will not be limited to one dimensional signals.