A wavelet filter comparison on multiple datasets for signal compression and denoising

• Published in: Multidimensional systems and signal processing Vol. 32; no. 2; pp. 791 - 820
• Main Authors: Gnutti, Alessandro, Guerrini, Fabrizio, Adami, Nicola, Migliorati, Pierangelo, Leonardi, Riccardo
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.04.2021; Springer Nature B.V
• Subjects: Artificial Intelligence; Circuits and Systems; Datasets; Electrical Engineering; Electrocardiography; Electromagnetic wave filters; Engineering; Noise reduction; Signal,Image and Speech Processing; Sound filters; Wavelet filter comparison; Discrete wavelet transform; Sub-band coding; Multiresolution analysis
• ISSN: 0923-6082; 1573-0824
• DOI: 10.1007/s11045-020-00753-w

In this paper, we explicitly analyze the performance effects of several orthogonal and bi-orthogonal wavelet families. For each family, we explore the impact of the filter order (length) and the decomposition depth in the multiresolution representation. In particular, two contexts of use are examined: compression and denoising. In both cases, the experiments are carried out on a large dataset of different signal kinds, including various image sets and 1D signals (audio, electrocardiogram and seismic). Results for all the considered wavelets are shown on each dataset. Collectively, the study suggests that a meticulous choice of wavelet parameters significantly alters the performance of the above mentioned tasks. To the best of authors’ knowledge, this work represents the most complete analysis and comparison between wavelet filters. Therefore, it represents a valuable benchmark for future works.