Speech Signal Analysis Using Fractal Geometry for Phoneme Boundary Detection

Phoneme segmentation deals with identifying the boundary points between two different phonetic units. It plays a vital role in automatic phoneme segmentation and annotating speech data. Deep learning methods require a huge amount of data for training which is not a favourable situation for low-resource languages and low-cost devices. Therefore, algorithms that require less data are desirable in both situations. In this paper, we propose a fractal-based method for phoneme segmentation that works by understanding the structural properties of a speech signal.