Darko Pekar

Papers

Total Citations

H-Index

About

Darko Pekar is a researcher advancing the frontier of expressive speech synthesis, with a focus on making human-robot interaction and digital communication more natural and engaging. His key research areas include neural network-based text-to-speech (TTS), speaker and style embedding, and multi-style voice adaptation for small datasets. Pekar’s most notable contribution is a novel architecture that uses speaker/style embeddings to train neural networks for synthesized speech in a specific voice and speaking style, even with limited target data—a breakthrough for personalized and adaptive voice interfaces. This work has garnered 11 citations, reflecting its relevance in the TTS community. He also compared multi-style DNN-based TTS approaches using small datasets (5 citations), demonstrating that people perceive interactions with computers and robots as social communication, underscoring the need for human-like speech. His research on expressive speech in human-robot interaction (4 citations) further explores how emotional and stylistic nuances can enhance user experience. Pekar’s work is pivotal for creating more intuitive, empathetic, and accessible voice technologies, bridging the gap between synthetic and human speech.