Portable implementation of a text-to-speech system for portuguese

Speech synthesizers are mostly developed for general purpose personal computers, which seems to be a problem for users who require mobility, as that is the case for disabled people. Therefore, a Portuguese speech synthesizing system inspired on the time domain pitch synchronous overlap and add technique was developed onto a digital signal processor due to its portability advantages. To allow full integration of the whole system, the original audio data, formed by 922 diphone units for the European Portuguese, was encoded using a code-excited linear prediction technique. By doing so, the database size dropped to less than 1% of its original size, with minimal loss in quality due to such coding process. The final system is able to operate with a personal digital assistant (PDA) or any custom device able to generate and send written text by means of the UART serial protocol.