Emotion-Aware Text to Speech: Bridging Sentiment Analysis and Voice Synthesis

This paper delves into the multifaceted realm of speech synthesis, focusing on challenges and opportunities in synthesizing speech from expressive corpora and mastering voice impersonation. Audiobooks, rich sources for speech synthesis, introduce variability in voice characteristics, necessitating nuanced approaches for unit-selection and statistical parametric speech synthesis. Voice impersonation, a complex phenomenon, involves intuitively mimicking another's voice, yet quantifying elusive elements poses challenges. Despite Text-to-Speech (TTS) advancements, expressive speech synthesis remains challenging, prompting exploration into emotions, speaking styles, and character voices. This study specifically concentrates on Emotion-Aware Text to Speech, emphasizing the usage of Hidden Markov Model (HMM) algorithm and the Hidden Markov Model Toolkit (HTK). Addressing voice variability, impersonation challenges, and expressive synthesis, the paper aims to enhance the authenticity of synthesized speech, advancing capabilities for human-computer interaction.