Abstract: In this paper, a new approach to identify ARMA model under noisy background is real-ized by using float-encoding genetic algorithms. With the aid of this method, unbiased parameters estimates of the system can be obtained when the output is corrupted by noise unrelated with the input signals. The effectiveness of the approach is testified by several simulations and the results show that it is more effective than least-square method. And it has been successfully applied in the measurement of the low-frequency characteristic parameters of a vented-box loudspeaker system.

Key words: system identification, ARMA model, float-encoding, genetic algorithm