Transmission Behaviour of the Human Speech Generating System in the Articulation of Different Short and Long Vowels

In earlier publications, Kiipfmuller and Cramer described a procedure for synthesizing speech from monophones by inserting a neutral transient between two successive sounds. Systematic investigations aimed at finding appropriate transients depending on the phonic environment showed that the sound character of several German vowels can be purposefully changed if the duration of their steady state is varied according to certain rules. This effect is particularly evident in the case of the vowel pairs /e:/, /I/; /ø:/, /Y/; /ε:/, /ε/ and /o:/, /U/ such as to allow both sounds of a pair to be interchanged with the aid of the procedure described.

In order to explain this phenomenon, experiments were made to find out whether the articulatory positions of the vocal tract are identical to the naturally spoken vowels under consideration. The transfer functions of the vocal tract for these vowels – spoken by a normal speaker who used, however, an artificial excitation of the speech generating system – were measured, with respect to magnitude and phase, according to different methods and compared. The measurements yielded largely identical amplitude and phase functions for the vowels of each pair. As a result of these tests, the differing auditive interpretation of the vowel character can be attributed to the fact that, in the perception of speech, vowels having the same spectral structure are especially evaluated with respect to their sound duration.