Andrew P. Lea, Quentin Summerfield, Minoru Tsuzaki, Alain de Cheveigné
THE PERCEPTION OF CONCURRENT VOWELS:
PERIODIC AND APERIODIC VOWELS
Abstract:The first aim of the experiments described in this paper was to determine whether listeners can use the differences between periodically and aperiodically excited speech to segregate concurrent vowels which consist of one periodic and one aperiodic vowel. The second aim was to establish whether listeners can segregate either the periodic constituent, the aperiodic constituent, or both constituents of the concurrent vowels. In Experiment 1, 2, 3, 4 and 6 listeners identified both constituents of pairs of vowels presented concurrently. The constituents were either pulse-excited with fundamental frequencies (f0s) of 100 Hz and 112 Hz, or noise-excited. The results of Experiments 1, 2, 3, 4 and 6 show that listeners can use the difference in voicing between aperiodic and periodic constituents to segregate them. These experiments also show that the ability of listeners to use this difference in voicing is as good as a difference in pitch between two voiced vowels. Experiment 1 shows that listeners segregate the periodic/aperiodic concurrent vowel by only segregating only the aperiodic constituent and not the periodic constituent. However, Experiment 2 shows that both the periodic and aperiodic constituents were segregated. A number of procedural differences existed between Experiments 1 and 2 which might explain these contrasting results. Experiments 3 and 4 removed the possibility of all but one of these differences explaining the contrasting results of Experiments 1 and 2. The remaining difference was the method of matching the amplitudes of the periodic and aperiodic vowels. Therefore, Experiment 5 conducted a loudness matching experiment to find the most appropriate method of matching the periodic and aperiodic vowels. Experiment 6 used the results of Experiment 5 to show that when the vowels are correctly matched for equal loudness both the periodic and aperiodic vowels can be segregated. The implications for competing speech segregation are discussed.