Title: Dictionary Learning Using EMD and Hilbert Transform for Sparse Modeling of Environmental Sounds

Year of Publication: Nov - 2014
Page Numbers: 104-110
Authors: Bochra Bouchhima , Rim Amara and Monia Turki-Hadj Alouane
Conference Name: The International Conference on Artificial Intelligence and Pattern Recognition (AIPR2014)
- Malaysia


This paper presents a new dictionary learning method for reconstruction tasks. Based on Empirical Mode Decomposition(EMD) and Hilbert Transformation, the dictionary is learnt, once and for all, from data having a complex structure and belonging to the same manifold. Searching sparse representation of signals over the learned dictionary is achieved by solving a minimum norm??least square error problem. Experiments on audio patterns demonstrate that it is effective. The dictionary is data adaptive and allows having signal sparse models which are well suited to restoration tasks. It has a good reconstructive sparse over??complete encoding and is largely competitive against KSVD and MOD algorithms. Our approach can be of much interest in off-line dictionary learning applications.