Rainer Gruhn, Harald Singer
Client-Server Speech Translation
System ATR-MATRIX
Abstract:We describe the implementation of a client-server-based speech translation system. To minimize the bandwidth problem, input speech data is preprocessed and compressed and then sent
to the recognition server. For synthesizing the translated utterance we implemented a variety of
approaches requiring between 256 kbps for high quality speech down to less than 1 kBps for unit
information. The client was implemented for Windows95/98, Linux and OSF1. The system was
also tested using a 9.6 kBps mobile telephone data connection. This report is accessible for ITL
members via /home/singer/tex/TR-IT-MATRIX/.