The aim of this project is to implement and validate the Fast ICA algorithm based on Negentropy, as described in the papers:
- Independent component analysis: An introduction. (Alaa Tharwat)
- Blind Source Separation of Underwater Acoustic Signal by Use of Negentropy-based Fast ICA Algorithm. (Tu Shijie and Chen Hang)
We reproduced the core functionalities of the algorithm and evaluated its performance using both synthetic data and simulated real-world scenarios. These scenarios utilized pre-recorded sound sources mixed in virtual environments created with the Pyroomacoustics library.
This experimental setup allowed us to separate and reconstruct the original sound sources using the FastICA algorithm and assess its effectiveness under various mixing conditions. We conducted the evaluation through direct comparisons and the application of Blind Source Separation (BSS) metrics.