Comparative Assessment of Wavelet and Kalman Filtering for Mode Estimation

This paper conducts a comparative assessment of Wavelet denoising and Kalman filtering as pre-processing tools for system identification techniques. Both tools are used to filter out noise from synthetic signals. The synthetic signals are distorted by varying signal-to-noise ratio levels and distributions of noise. The effectiveness of the examined methods is evaluated by identifying the oscillatory modes contained in the synthetic signals. Mode identification is performed using a hybrid frequency-domain/time-domain (FD/TD) approach. Results reveal a high level of accuracy for all noise conditions, with Wavelet denoising method exhibiting a slightly better performance compared to the Kalman filtering technique. The latter is particularly true under highly noisy conditions.