الفهرس | Only 14 pages are availabe for public view |
Abstract Noise is a major problem in recognition field. It could degrade the quality of communication, cause transmission errors and may even disrupt a measurement process. Hence the methods to remove noise become most significant in signal analysis. This thesis is aimed to remove the noise from radar signal by recognizing it. Machine learning could recognize the noise in part of second. By knowing the kind of noise it become easy to choose the best filter to remove it In this thesis four different types of noises are recognized they include gamma noise, Chi - square noise, uniform noise and white gaussian noise. All experiment results presented in the thesis are based on both synthetic data from MATLAB and real data from Jim Lux website. Five featured are extracted to define these noises. The features are mean, variance, skweness, kurtosis and range. Neural network architecture, discriminate analysis, single decision tree, support victor machine and K-means cluster were also used to classify the noise and their performances were compared The result showed that the Discriminate Analysis was the best classify comparing with other. Which had a highest accuracy with shortest time |