Project for machine learning class at Northwestern University. For more info check out the project website (http://crxpandion.github.com/Singer_ID/)
This is project contains several scripts and programs to identify interesting characterisitcs in signers' voices.
First, a wav file is loaded into matlab and processed into mel frequency cepstral coefficients and their composite transfer function is estimated using an auto regressive filter. Then we output this data into a csv file and run a quadratic SVM on this data.
We utilize LibSVM and Dan Ellis' melfcc calculator.