Identification of Synthetic Isolated Vowels and Vowels in H-D Context

During the course of investigations into the effects of various acoustic parameters on the intelligibility of synthesised vowels it was observed that many listeners were able to identify vowels more reliably and uniformly when the sounds were embodied in h-d words than when vowels were presented in isolation. The sounds which were to be identified were acoustically identical in both instances. This paper describes experiments in which various acoustic features of the / h / and / d / were systematically eliminated in order to locate which parts of the consonants aid this more uniform identification. It is concluded that no one consonantal feature was responsible but that a spontaneous improvement in uniformity of identification occurs when several contextual features combine to achieve a new level of speech-likeness.

This paper is concerned with a more detailed examination of the h-d context in an attempt to isolate characteristics which contribute most to the result of our earlier experiment [3].