Telugu emotional story speech synthesis using SABLE markup language

Abstract

In this paper, a framework for synthesizing Telugu emotional speech for story telling applications is presented. An XML based markup langauge, SABLE is used to synthesize the emotions from a given story text. SABLE markup defines a set of tags to improve the quality of the synthesized speech from the concatinative speech synthesizer. In this work, a subset of prosody tags are used to synthesize the emotional speech from a given story text. Modified Zero frequency filtered (ZFF) signal is used to derive the prosody correlates of pitch base, pitch range and intensity. The desired prosody modification factors for each emotion is derived at phrase level. The derived prosody modification parameters for each emotion are stored in the form of a template. During synthesis, hand annotated story text is replaced by prosody tags which are stored in templates. Prosody tagged story text at phrase level is automatically converted into SABLE markup format. The markup story text is used to synthesize emotional speech from the Telugu neutral Festival TTS system. The quality and naturalness of the synthesised emotional story speech is evaluated using subjective tests.