The Cocktail Party
The objective of this project was to separate mixed audio signals originating from different sources, allowing for the isolation of each individual signal. To achieve this, I implemented the Independent Component Analysis (ICA) algorithm, as described by Hyvärinen et al. (2001). This work was completed as part of the Digital Audio INF8703 course at Polytechnique Montréal. An example is provided below to demonstrate the results. In this example, the voices of Hillary Clinton and Donald Trump were mixed using different coefficients to simulate distinct sources. The ICA algorithm successfully separated these mixed signals into two individual signals, each corresponding to one of the voices.
