Even though Windows CE devices gain more and more popularity there are still not nearly as many software packages available for these machines as there are for platforms like DOS, Windows 95 or NT. This is especially the case for public domain software. telos intends to develop some CE applications in the future which will use speech output. Our original idea was to base these applications on a freely available text-to-speech library for windows CE. However, we learned that such library doesn’t seem to exist.

We therefore decided to port an existing free text-to-speech system to Windows CE.

The RSynth package was created and is distributed by Nick Ing-Simmons under UNIX and there is a port for Win32. Windows 95, NT and UNIX are all workstation platforms with plenty of both main and disk memory. On windows CE machines there is usually no hard disk installed and the available RAM is quite limited.

Therefore porting the RSynth software to this type of computer will mean a rewrite of many portions of the code. RSynth utilises dictionaries to improve speech quality. Such dictionaries are not a choice for CE since they would use up almost all of the memory which is usually available for file storage on typical devices. Therefore our intention is to improve the rules implemented in RSynth for words which are not found in the dictionary.

Another consideration is the performance. CE machines usually do not have a numeric coprocessor and therefore arithmetic operations which are widely used in RSynth take quite some time. We therefore need to optimise on the amount of arithmetic used in the process of speech generation.

The goal of the CE speech project is to obtain a library which can be easily interfaced by applications to convert plain text into synthesised speech. After the project has advanced to a presentable state we will provide the results on our web site and to the original author of RSynth.

Sorry, but since 2003 this project is not supported anymore.