A Detailed Subjective Evaluation of Devanagari Based Text To Speech System

Autor: Ruchi Raheja, Swaran Lata, Ankit Kesarwani, Somanath Chandra, Mahesh Kulkarni
Rok vydání: 2018
Předmět:
Zdroj: 2018 3rd International Conference On Internet of Things: Smart Innovation and Usages (IoT-SIU).
DOI: 10.1109/iot-siu.2018.8519838
Popis: This paper is about the detailed subjective evaluation of Indic Text to Speech system for Hindi and Marathi languages which are based on Devanagari script. In this evaluation, various speech quality aspects, namely naturalness, intelligibility, pronunciation and comprehensibility along with language specific parameters, are evaluated. And this can be considered as an exhaustive evaluation in which all major speech parameters have been observed in detail. To evaluate these parameters, a comprehensive test suite and an evaluation assistance application have been designed and developed. This kind of detailed evaluation has been carried out on Hindi and Marathi language for the first time, which has resulted in concluding the TTS as an acceptable system by the end user with some improvement areas wherein researchers still needs to focus. Observations and results of various approaches are also demonstrated in this paper.
Databáze: OpenAIRE