Related Experiment Videos
On neural blind separation with noise suppression and redundancy reduction
J Karhunen1, A Cichocki, W Kasprzak
1Helsinki University of Technology, Rakentajanaukio, Finland. Juha.Karhunen@hut.fi
International Journal of Neural Systems
|April 1, 1997
Summary
This study introduces new methods to reduce noise in blind source separation (BSS) for sensor signals. The research focuses on scenarios with more sensors than sources, demonstrating effective noise elimination.
Area of Science:
- Signal Processing
- Machine Learning
- Data Science
Background:
- Real-world sensor signals are inherently affected by additive and convolutive noise.
- Blind Source Separation (BSS) aims to recover original signals from mixed observations.
- Existing BSS methods face challenges, especially in underdetermined or overdetermined scenarios.
Purpose of the Study:
- To investigate the reduction and elimination of additive and convolutive noise in BSS.
- To develop novel methods for the extended BSS problem where the number of sensors exceeds the number of sources.
- To propose adaptive learning algorithms for these advanced BSS scenarios.
Main Methods:
- Development of various signal processing techniques for noise reduction in BSS.
- Formulation of adaptive learning algorithms tailored for overdetermined BSS.
- Utilizing computer simulations to validate the proposed methodologies.
Main Results:
- Demonstrated significant reduction and potential elimination of both additive and convolutive noise.
- Successfully addressed the extended BSS problem with more sensors than sources.
- Validated the effectiveness and performance of the proposed methods through extensive simulations.
Conclusions:
- The proposed methods offer effective solutions for noise reduction in BSS.
- The study provides viable approaches for overdetermined BSS scenarios.
- Computer simulations confirm the practical applicability and performance of the developed algorithms.