Investigations into prosodic syllable contour features for speaker recognition