Application of higher-order spectra in on-line identification for ANC systems

This paper deals with the problem of electro-acoustic plant models identification during the operation of active noise control system. For this purpose, the identification method employing higher-order spectra aided by signal averaging procedure is proposed. This novel approach improves the quality of identified models in comparison to models obtained via conventional methods based on second-order spectra (i.e. power spectral densities). The proposed method allows for reducing the influence of additive Gaussian disturbance on identification results, because higher-order spectra of Gaussian processes are identically equal zero. Additionally, repetition of excitation sequence and averaging of signals allow for enhancing signal-to-noise ratio without deteriorating noise attenuation. The accuracy of obtained models and their influence on the attenuation level is tested in different simulation experiments, imitating the operation of real-world laboratory active noise control system.