Development of ECA-based text-to-speech synthesis system PLATTOS for conversational intelligent systems

TTS PLATTOS: Evaluation

Listening tests are often used to assess the quality of the audio system. In recent years such tests were performed mostly in controlled environments with selected participants and professional audio equipment. In the last few years, conducting listening tests via the Internet, or so-called web-based experiments, has become very popular. Namely, the studies showed that web-based experiments provide comparable results as laboratory experiments. Until recently, only a limited number of types of listening tests could be conducted in the form of web-based experiments because some key features were missing in web standards, such as sample manipulation of audio stream patterns. With the advent of the Web Audio API, however, a much wider range of listening tests has been made possible, as new audio processing functionalities have been supported. In the case of text-to-speech synthesis systems, ITU-R BS.1534 (MUSHRA) is recommended for performing listening tests.

• MUSHRA-learning evaluation

TTS PLATTOS: architecture-0

TTS PLATTOS: architecture-2

TTS PLATTOS: architecture-3

TTS PLATTOS: architecture-4