Classification of real time moving object using echo state network

It is well-known that an artificial agent exhibits deterministic dynamics when it moves in a closed real world environment. It is interesting to determine this dynamics when a real biological being such as fish is kept in a real closed environment and free to move in it. This paper determines some deterministic dynamics of fish motion freely moving in a closed environment. The task is divided into several stages - image capturing, image processing, time series extraction, Chaos analysis and Classification. The classification performance is analyzed with feed forward neural network (FNN), Recurrent Neural network, Fuzzy network, Bagged Regression trees and Echo state network (ESN). Simulation result exhibits that the proposed ESN algorithm outperforms other networks.