presented at event 2010 IEEE International Conference on Acoustics, Speech and Signal Processing Conference