---

This work investigates one of the most attractive research areas of blind signal decomposition, performed with very little a priori knowledge about the mixing model or sources characteristics, with the only knowledge of the observed mixed signal. Blind Source Separation (BSS) is a method to separate a target signal from mixtures by means of some modern statistical component analysis techniques, like Independent Component Analysis (ICA), Sparse Code (SC) and Non-negative Matrix Factorization (NMF) which differentiate each other for the imposed mathematical assumptions (or constraints). The goal of this work is to develop signal decomposition applications based on the BSS approach, in single-channel strategy, like the audio restoration of 78 rpm gramophone discs or in multi-channel strategy, like the estimation of the original source waveforms from a seismometer array near an active volcano. Although the BSS approach realizes the powerful task of signal decomposition and features extraction, considerable work is needed to model and implement the complete framework.

---

Giuseppe Cabras is IT Technician at the Department of Chemistry Physics and the Environment, associate member of InterUniversity Center for Behavioral Neurosciences (ICBN) and received his PhD degree in Industrial and Information Engineering at the Department of Electrical, Management and Mechanical Engineering of the University of Udine.

---