Enhancing Indonesian Speech Synthesis: Embracing Naturalness and Expressiveness with Hidden Markov Models

This study employs a pure experimental research approach, investigating speech-based interfaces with a focus on speech recognition and synthesis (text-to-speech). The primary goal is to create natural conversational interfaces. The methodology involves integrating data collection, speech signal analysis, and speech synthetic design, encompassing system analysis, design, and implementation. Through iterative prototype testing, followed by rigorous pre-and post-tests, the research ensures the application’s practical readiness. Incorporating monophone-based speech synthesis enhances adaptability to diverse speech traits and enables expressive synthesis. Utilizing the Hidden Markov Model method, the research aims to translate written text into authentic Indonesian speech, emphasizing naturalness in the synthesized output. This experimental design systematically seeks an optimized interface for seamless and authentic Indonesian speech synthesis from text. The research demonstrates the effectiveness of the Hidden Markov Model in translating text into natural Indonesian speech. Furthermore, achieving natural speech articulation and improved acoustic models requires analyzing phonology, linguistic information, and speech structure, which are crucial in constructing contextual factors during system development.