Perception of phase in the glottal excitation and its relevance in statistical parametric speech synthesis

Tuomo Raitio, Lauri Juvela, Antti Suni, Martti Vainio, and Paavo Alku
Submitted to Speech Communication: Special Issue on Phase-Aware Signal Processing



Sound examples of natural speech (experiment 1):

method/
sample
Natural phase Zero-phase Cyclostationary-random-phase Random-phase
M1 95 Hz
M2 113 Hz
M3 202 Hz
M4 85 Hz
M5 90 Hz
M6 242 Hz
F1 258 Hz
F2 129 Hz
F3 220 Hz
F4 152 Hz
F5 378 Hz




Sound examples of vocoded speech (experiment 2):

method/
sample
Natural speech Vocoded natural phase Vocoded zero-phase Vocoded cyclostationary-random-phase
M1 breathy
M1 normal
M1 Lombard
M2 breathy
M2 normal
M2 Lombard
M3 breathy
M3 normal
M3 Lombard
F1 breathy
F1 normal
F1 Lombard
F2 breathy
F2 normal
F2 Lombard




Sound examples of HMM-based speech synthesis (experiment 3):

method/
sample
Synthetic natural phase Synthetic zero-phase Synthetic cyclostationary-random-phase
M1 breathy
M1 normal
M1 Lombard
M3 breathy
M3 normal
M3 Lombard
F1 breathy
F1 normal
F1 Lombard
F2 breathy
F2 normal
F2 Lombard




Repeated measures ANOVA results for experiment 1


Repeated measures ANOVA results for experiment 2


Repeated measures ANOVA results for experiment 3



Last updated 15.6.2015.