SYNTEX - A microprocessor based system for automatic conversion of German text to speech

Abstract

SYNTEX has been developed for real time conversion of unrestricted German text to speech. Main part of the system is a word parser used to decompose compounds, to mark affixes, and to detect accented syllables in foreign words. The results of this parsing are used by a set of letter—to—sound rules working on morhs and by algorithm for computation of word intonation. The intonation algorithm treats German words as well as foreign ad.aptions. It assigns main and rhythmical stress and converts it to a fundamental frequency contour. SYNTEX is supported by a text preprocessor treating upper and lower case letters, numbers and punctuation. Routines for the conversion of phonemes and intonation contour to synthesizer parameters are written for a Votrax VS6-G2 so far, but can easily be adapted to other synthesizers. THE CONCEPT OF SYNTEX SYNTEX has been planned as a handy and inexpensive device to be used for example as a tool for the handicapped. When work started, commercial phoneme synthesizers were already available, consequently it was decided to use a commercial synthesizer and to concentrate on a system to turn text into synthesizer parameters. The concept of this system is shown in Fig. 1. SYNTEX is running on a microprocessor system (using a Motorola M 6800) and needs about 10 K byte of memory. As a synthesizer, the Votrax VS6-G2 is used, but SYNTEX can be adapted to other synthesizers easily. In order to process speech in real time, conversions are done by rules and without a dictionary. Though parts of the concept are still missing, about 15 words are processed in the time used to speak one word. At this time, SYNTEX has a structure as shown in Fig. 2, consisting of text preprocessing, word parsing, letter—to— sound rules and word intonation rules. OH 1746•7!821000

DOI: 10.1109/ICASSP.1982.1171809

2 Figures and Tables

Cite this paper

@inproceedings{Rhl1982SYNTEXA, title={SYNTEX - A microprocessor based system for automatic conversion of German text to speech}, author={Hans-Wilhelm R{\"{u}hl}, booktitle={ICASSP}, year={1982} }