Dissertation in the field of Speech and language technology, Tuomo Raitio

The title of thesis is Voice source modelling techniques for statistical parametric speech synthesis

Speech is the most natural way through which humans communicate, and today speech synthesis is utilised in various applications. However, the performance of modern speech synthesisers falls far short from the abilities of human speaker— synthesising intelligible and natural sounding speech with desired contextual and speaker characteristics and appropriate speaking style is extremely difficult. This thesis aims to improve both the naturalness and expressivity of speech synthesis by proposing new methods for voice source modelling in statistical parametric speech synthesis. With accurate estimation and appropriate modelling of the voice source signal, which is known to be the origin for several essential acoustic cues in spoken communication, various expressive voices are created with high degree of naturalness and intelligibility. Evaluations in various listening contexts show that speech created with the proposed methods is assessed to be more suitable than that generated with current techniques, thus providing potentially large benefits in many speech synthesis applications.

Opponent: Professor Yannis Stylianou, University of Crete, Greece

Supervisor: Professor Paavo Alku Aalto University School of Electrical Engineering, Department of Signal Processing and Acoustics

Dissertation website

Contact information
Tuomo Raitio
p. 0440 668 669
p. +1 408 858 9068
tuomo.raitio@aalto.fi