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