In the talk, I will give an overview of methods for audio source separation based on non-negative matrix factorization as well as several recent extensions thereof.