Speech Processing by Computer

 

LECTURE 8

SIGNAL GENERATION FOR SPEECH SYNTHESIS

 

Objectives

 

By the end of the session you should be able to:

·        describe the basic operation of concatenative signal generation systems

·        describe in general terms the signal processing requirements of concatenative systems

·        describe the basic operation of formant synthesis systems

·        discuss the relative advantages and disadvantages of concatenative and formant techniques for signal generation.

 

Outline


 

8.      Signal Generation

8.1.   Requirements

8.2.   Concatenative Synthesis

8.2.1.      Diphones

8.2.2.      Other units

8.2.3.      Corpus based

8.2.4.      Signal Processing

8.2.4.1.Duration modification

8.2.4.2.Pitch modification

8.2.4.3.Smoothing at joins

8.3.   Formant Synthesis

8.3.1.      Architecture

8.3.1.1.Series vs. parallel

8.3.2.      Parameter generation

8.4.   Articulatory Synthesis

8.4.1.      Aims & problems

8.4.2.      High-level synthesis

           

Readings

 

General:

            T. Dutoit, “Synthesis Strategies”, in An introduction to text-to-speech synthesis, Kluwer 1997.

 

Specific:

            P. Bhaskararo, "Subphonemic segment inventories for concatenative speech synthesis", in Fundamentals of Speech Synthesis and Speech Recognition, ed Eric Keller, Wiley, 1994.

 

            C. Hamon, E. Moulines, F. Charpentier, "A Diphone Synthesis System based on Time-Domain Prosodic Modifications of Speech", Proc. Int. Conf. Acoustics Speech and Signal Processing, 1989,  pp238-241.

 

            D. Klatt, “Software for a cascade/parallel formant synthesizer”, J.Acoust.Soc.Am 67(3) 1980, 971.