Design, implementation and evaluation of a scanned Filipino text-to-speech device as a reading aid for the blind and visually impaired

• Main Authors: De Leon, Christian Paulo L., Gatchalian, Allan Adrian B., Ninobla, Juan Miguel D., Concepcion, Nigel Roi D., Catemprate, David Joseph O., Pascual, Ronald M.
• 格式: text
• 出版: Animo Repository 2018
• 主題: Text-to-speech software; Computer Sciences
• 在線閱讀: https://animorepository.dlsu.edu.ph/faculty_research/5764
• 機構: De La Salle University

In this study, the authors provide new insights to the design, implementation and evaluation of a TTS system where the inputs are text images in Filipino. Many text-to-speech (TTS) systems have been studied and implemented in many languages. Portable devices for Filipino TTS however have not been explored yet. With the aid of portable scanning device and Optical Character Recognition (OCR), our TTS system can generate synthetic Filipino speech from texts found on printed reading materials in Filipino. The authors present the development of a scanned Filipino TTS converter device as a reading aid for the blind and visually impaired. The authors implemented the design using a raspberry pi as its main control unit, and interfacing it with a text scanner, a speaker system, and a power supply module. The system uses 3000-word database adapted from the most commonly used words in the Filipino speech corpus from UP – Sentro ng Wikang Filipino. Employing a certain speech synthesis toolkit [8], the synthesized Filipino words were generated by creating phonetic records that emulate the Filipino speech. Listening tests were conducted to measure the quality of the synthesized Filipino speech. The aforementioned tests involved 60 respondents rating the quality of the output speech. Results showed that overall, on a scale of 1 to 5 with 5 being the highest, the respondents rated the quality of the output speech at 4.1333.