Wimol San-Um

pp. 699 - 674

Abstract

This paper presents a robust cellular associative memory for pattern recognitions using composite trigonometric
chaotic neuron models. Robust chaotic neurons are designed through a scan of positive Lyapunov Exponent (LE) bifurcation structures, which indicate the quantitative measure of chaoticity for one-dimensional discrete-time dynamical systems.
The proposed chaotic neuron model is a composite of sine and cosine chaotic maps, which are independent from the output
activation function. Dynamics behaviors are demonstrated through bifurcation diagrams and LE-based bifurcation structures.
An application to associative memories of binary patterns in Cellular Neural Networks (CNN) topology is demonstrated using
a signum output activation function. Examples of English alphabets are stored using symmetric auto-associative matrix of
n-binary patterns. Simulation results have demonstrated that the cellular neural network can quickly and effectively restore
the distorted pattern to expected information. 

Keywords

• associative memory
• pattern recognition
• trigonometric chaotic neuron