Multilingual Text-to-Speech Software Component for Dynamic Language Identification and Voice Switching

Text-to-speech synthesis is a critical feature of the applications developed for people with visual or reading disabilities. In the last years there has been an increasing interest in multilingual text-to-speech synthesis, which requires multilingual text analysis and language specific speech synthesis. In this case, the dynamic switching of the synthetic voice is needed in order to enhance the usability and user experience. This paper aims at presenting a software component for multilingual text-to-speech synthesis. The software has been developed and tested in four steps: alpha version (proof-of-concept), functional version (beta), commercial version, and implementation. The beta testing results showed a high accuracy of the language detection algorithms, which perform properly on texts having a variable degree of fragmentation. The commercial version has been then successfully implemented in two applications for visually impaired people: an automatic reading machine and a personal organizer for the blind and visually impaired users. Both implementations have been tested with users for usability and acceptance. The evaluation results showed that a device with this component is easier to use by visually impaired people.