Nonlinear Adaptive Noise Cancellation for 2-D Signals with Adaptive Neuro-Fuzzy Inference Systems

Nonlinear Adaptive Noise Cancellation for 2-D Signals with Adaptive Neuro-Fuzzy Inference Systems Hao Qin Advisor: University of Guelph, 2004 Professor Simon X. Yang Neuro-fuzzy systems are capable of inducing rules from observations, where the adaptive neuro-fuzzy inference system (ANFIS) is an effective method that can be applied to a variety of domains such as pattern recognition, robotics, nonlinear regression, nonlinear system identification, and adaptive signal processing. However, the signals processed by a typical ANFIS are in one dimension, e.g., acoustic signals. In this thesis, we extend the ANFIS method to two-dimensional (2-D) signals. First, the 2-D signal restoration contaminated with Gaussian noise is investigated in nonlinear passage dynamics of orders 2 and 3. We inspect eight types of membership functions (MFs): bell MF, triangle MF, Gaussian MF, two sided MF, pi-shaped MF, product of two sigmoidal MFs, difference of two sigmoidal MFs, and trapezoidal MF. In addition, several other parameters, such as the training epochs, the number of membership functions for each input, the optimization method, the type of output membership functions, and the over-fitting problem, are investigated. Secondly, ANFIS is used to restore the 2-D signals corrupted by salt and pepper noise. Eight types of membership functions and several other parameters are presented. We also compare the effect of the 2-D signal restored by ANFIS with the conventional filters including spatial filters, frequency domain filters, adaptive optimal filter, Wiener filter, and wavelet and wavelet packet.