Kalman Filter as a pre-processing technique to improve the support vector machine

The Support Vector Machine is widely used as a classification tool to analyze data and recognize patterns. In certain applications of Support Vector Machine, noisy data can greatly affect accuracy and performance. To improve the accuracy of the system, the Kalman Filter has been proposed as a suitable pre-processing technique which can be implemented before using the Support Vector Machine to classify the information. This system has been tested using a dataset obtained from a pipeline defect monitoring system in the department´s pipeline laboratory. This test rig uses long range ultrasonic testing to detect minor defects inside a stainless steel pipe. MATLAB simulations show promising results where Kalman Filter and Support Vector Machine combination in a single system produced higher accuracy compared to the discrete wavelet transform in a noisy environment.