Animal Vocalization Analysis and Synthesis

This Master’s thesis focuses on the unexplored domain of animal vocalizations. In the present work Lemur (madagascarian mammal) vocalizations are studied and analysed in order to create a system that can process and train datasets and posteriorly synthesize vocalizations. To achieve the goal the technique of Hidden Markov Models based synthesis is used with the HMM-based speech synthesis system (HTS) toolbox, based on the wide known Hidden Markov Model Toolkit (HTK) speech recognition system. In the following pages a system to synthesize lemur vocalizations is presented and its advantages and inconveniences defined.