Speaker: Juan Ignacio Álvarez Trejos.
Abstract: X-Vectors are speaker embeddings that emerge to address the speaker recognition task, surprisingly outperforming i-vectors in most speaker tasks. It is proposed to take advantage of the information contained in these embeddings by using them in the speaker diarization task. For this purpose, the Self Attentive-End to End Neural Diarization (SA-EEND) model will be adapted to introduce X-Vectors and thus even improve the information contained in them. Several systems will be presented by the AUDIAS group to the Albayzin 2022 evaluation for the tasks of diarization and identity assignment, including two of those presented in this talk.