Magnitude-aware Probabilistic Speaker Embeddings
This work addresses speaker recognition for applications like security and communication by improving accuracy through quality-aware methods, but it is incremental as it builds on existing hyperspherical embedding techniques.
The paper tackled the problem of ignoring embedding magnitudes in speaker recognition by proposing a probabilistic speaker embedding extractor that leverages magnitude information for quality assessment and out-of-distribution detection, resulting in significant improvements over magnitude-agnostic baselines in speaker verification and diarization tasks.
Recently, hyperspherical embeddings have established themselves as a dominant technique for face and voice recognition. Specifically, Euclidean space vector embeddings are learned to encode person-specific information in their direction while ignoring the magnitude. However, recent studies have shown that the magnitudes of the embeddings extracted by deep neural networks may indicate the quality of the corresponding inputs. This paper explores the properties of the magnitudes of the embeddings related to quality assessment and out-of-distribution detection. We propose a new probabilistic speaker embedding extractor using the information encoded in the embedding magnitude and leverage it in the speaker verification pipeline. We also propose several quality-aware diarization methods and incorporate the magnitudes in those. Our results indicate significant improvements over magnitude-agnostic baselines both in speaker verification and diarization tasks.