Robust acoustic domain identification with its application to speaker diarization

With the rise in multimedia content over years, more variety is observed recording environments of audio. An audio processing system might benefit when it has a module to identify acoustic domain at its front-end. In this paper, we demonstrate idea identification (ADI) for speaker diarization. For this, first present detailed study various domains third DIHARD challenge highlighting factors that differentiated them from each other. Our main contribution develop simple and efficient solution ADI. work, explore embeddings task. Next, integrate ADI with diarization framework III challenge. The performance substantially improved baseline thresholds agglomerative hierarchical clustering were optimized according respective domains. We achieved relative improvement than $$5\%$$ $$8\%$$ DER core full conditions, respectively, on Track 1 evaluation set.