Computer Science Department
Blind Source Number Separation Estimation
Philip Vander Broek
This project is a very specific, research oriented topic that attempts to tackle Blind Source Separation (BSS). BSS is the separation of a set of signals from a set of mixed signals, without the aid of information (or with very little information) about the source signals or the mixing process. The over determined case of BSS simply means there are more microphones than there are original sources. Here is an example: There are 5 microphones randomly placed in a room, and x (less than 5) people randomly placed in that room. The people, or original sources, are recorded talking which produces the mixed microphone signals. The microphone signals will be audibly mixed, or contain multiple voices (in other words, the microphones are not directional; they simply record ambient sounds of the room). Without ever knowing the number x, by analyzing the 5 microphone signals it may be possible to estimate the number x. This project is essentially to create a computer program to estimate the number of original sources. The program will be developed to read in multiple microphone signals consisting of mixed audio sources, analyze the signals and, ideally, return the correct number of original sources. This primarily involves 1) researching and developing an algorithm to analyze the signals and 2) the creation of sample data to test the program.