This book investigates the challenge of speaker identification in scenarios where labeled training data is scarce or unavailable. It explores advanced machine learning techniques, particularly focusing on unsupervised and semi-supervised methods, to enable systems to learn speaker characteristics from unlabeled or minimally labeled audio streams. The work aims to advance the state-of-the-art in automatic speaker recognition by developing robust algorithms that can adapt to new speakers and acoustic conditions with limited prior information.

This work is positioned as a technical monograph for researchers and practitioners in the field of speech processing and machine learning. Its focus on self-learning techniques suggests a deep dive into algorithmic development and theoretical underpinnings, likely appealing to a specialized audience. The book's contribution is expected to be in advancing the methodologies for speaker identification under data-scarce conditions, offering new tools and insights for the academic and industrial research communities.