Omilia R&D team is working on novel methods for improving speech and speaker recognition technologies. In this paper, the team is collaborating with Brno University of Technology (Czechia) in order to approach the problem of speaker diarization using novel probabilistic speaker representations. The paper will be presented at Odyssey 2020 in Tokyo (Japan) in October, and we are happy to share the preprint version. https://omilia.com/wp-content/uploads/2020/04/Probabilistic-embeddings-for-speaker-diarization.pdf
Anna Silnova, Niko Brummer, Johan Rohdin, Themos Stafylakis, Lukas Burget
Brno University of Technology, FIT, IT4I CoE, Czechia
Omilia Conversational Intelligence, Athens, Greece
Abstract
Speaker embeddings (x-vectors) extracted from very short segments of speech have recently been shown to give competitive performance in speaker diarization. We generalize this recipe by extracting from each speech segment, in parallel with the x-vector, also a diagonal precision matrix, thus providing a path for the propagation of information about the quality of the speech segment into a PLDA scoring backend. These precisions quantify the uncertainty about what the values of the embeddings might have been if they had been extracted from high quality speech segments. The proposed probabilistic embeddings (x-vectors with precisions) are interfaced with the PLDA model by treating the x-vectors as hidden variables and marginalizing them out. We apply the proposed probabilistic embeddings as input to an agglomerative hierarchical clustering (AHC) algorithm to do diarization in the DIHARD’19 evaluation set. We compute the full PLDA likelihood ‘by the book’ for each clustering hypothesis that is considered by AHC. We do joint discriminative training of the PLDA parameters and of the probabilistic x-vector extractor. We demonstrate accuracy gains relative to a baseline AHC algorithm, applied to traditional x-vectors (without uncertainty), and which uses averaging of binary log-likelihood-ratios, rather than by-the-book scoring.
View the full paper here.