A Comparative Re-Assessment of Feature Extractors for Deep Speaker Embeddings
This work addresses the need for improved feature extraction in speaker verification, but it is incremental as it re-assesses existing methods rather than introducing new ones.
The paper tackled the problem of limited exploration of alternative feature extractors for deep speaker verification by re-assessing 14 methods on VoxCeleb and SITW datasets, finding that features like spectral centroids and group delay function reduced equal error rates by up to 16.3% and 25.1% compared to the baseline.
Modern automatic speaker verification relies largely on deep neural networks (DNNs) trained on mel-frequency cepstral coefficient (MFCC) features. While there are alternative feature extraction methods based on phase, prosody and long-term temporal operations, they have not been extensively studied with DNN-based methods. We aim to fill this gap by providing extensive re-assessment of 14 feature extractors on VoxCeleb and SITW datasets. Our findings reveal that features equipped with techniques such as spectral centroids, group delay function, and integrated noise suppression provide promising alternatives to MFCCs for deep speaker embeddings extraction. Experimental results demonstrate up to 16.3\% (VoxCeleb) and 25.1\% (SITW) relative decrease in equal error rate (EER) to the baseline.