Adaptive filters embedded in a user-friendly virtual instrument (VI) for signal denoising are proposed in the paper. The filters are based on the local polynomial approximation (LPA) and equipped with the improved intersection of confidence intervals (ICI) rule, known as the relative intersection of confidence intervals (RICI) rule, in order to achieve adaptivity to high transitions in signals and to vary the estimator size. The VI is also equipped with multiwindow estimation which significantly improves denoising quality, while preserving edges, instantaneous slope changes, and spikes in the signal. To reduce computational time, the fast ICI and fast RICI methods are provided. The LPA-RICI method is shown to outperform other adaptive denoising methods presented in the paper, reducing the estimation mean absolute error by up to 51.6 %, and the mean squared error by up to 81 %, while enhancing the execution time by up to 2 % when compared to the original LPA-ICI method.
The research was conducted at the Faculty of Engineering, University of Rijeka (www.riteh.uniri.hr).