Speech Processing

Mobile Applications

Keywords

Speech synthesis diphones,mobile phone, face animation, detection

Abstract:
This project is devoted to speech synthesis, face animation and development of an application for reading short messages in cell phones. Together, as seen in short demo, they are a part of a more complex project for multimedia reading of SMS on the mobile phone. After receiving the SMS a talking face based on a sender’s photo will appear on the screen and animate the reading while the speech will be synthesized parallely. This work includes analysis of implementation of the speech synthesizer – this means loading the database, synthesis, creating the annotation file and creating the output sound signal. Annotation file (*.ano) is needed for synchronization with the speaking face and it is a simple text file containing phonemes with its time marks (boundaries). Application also needs deformation of the general 3D model which is based on face features detection and detected features are used also for creation of texture. The final synthesized speech utterance is played together using the previous parts - and the result is face animation of talking human face. The type of created application is Java Midlet, very similar to applet class, but assigned for cell phones. Application is installed from JAR file – an archive that contains compiled classes. The application can be very easily modified for needs of a mobile phone. This will be necessary for older types of mobile phones, which have limited memory for applications - in this case the database is loaded like external file, and it’s not part of the JAR file. However, this change can also be very practical – if there is some change in database (like uved priklad ak chces) – it’s not necessary to change JAR file.

You can download a sample here