Comparison of system identification methods on movement dynamics of weakly electric fish

Purpose: The aim of this study is to compare eight different system identification methods on a biological system. In this study, the motion dynamics of the refuge-tracking behaviour of weakly electric fish were used as a dataset. Theory and Methods: Frequency Response Function, Recursive Least Squares Method, Spectral Analysis Method and WienerHopf Method were used in this study. ARX, ARMAX and Nonlinear Hammerstein-Wiener methods were used as parametric methods. As a subspace identification method, the n4sid method, which is widely used in the literature, was preferred. Results: The results obtained show that despite various differences and disadvantages, the estimation performance of all methods gives good results. Methods such as Frequency Response Function, ARX and ARMAX Models and N4SID gave better results in general. Conclusion: Compared to the alternatives, subspace identification methods stand out due to their ability to eliminate noise well and to express even complex systems with low-dimensional linear systems. Especially in studies on complex systems such as living organisms, the use of subspace identification methods will help to produce more effective results.