presented at event Proceedings of International Conference on Acoustics, Speech and Signal Processing (ICASSP-95), Detroit, MI Conference