Sound Source Localization and Separation

People face the problem of sound source localization and separation in situations where they attempt to localize and focus on a source of sound among a dissonance of conversations and background noise. This paper synthesizes a sound source localization routine. We utilize a general source separation technique, Independent Component Analysis.. Particularly, basic ICA was applied to separate mixtures of low frequency, narrow band, non-Gaussian signals by using closely spaced uni-directional microphones. The localization routine worked with an average condition number of 10. The routine was tested on data collected in a laboratory. 4 Acknowledgements I would like to extend my sincere gratitude towards Professor Daniel Kaplan and Professor James Heyman who provided guidance and supervision throughout this project. I also would like to thank the Macalester Keck Foundation, which has funded my research in Summer 2005. I thank my parents, Tesfaye and Fetlework, whose life story instilled a positive energy and a sense of responsibility in my soul. I am very grateful to the efforts of my siblings who home schooled me as I struggled with child-sickness, which hindered my regular school attendance: my eldest brother, Zelshe, taught me how to write; my brother Muluye and my sister Menbi (Dmet) taught me how to count. This Honors Thesis in Computer Science would never be realized without the help of all these people.