Partial AUC optimization based deep speaker embeddings with class-center learning for text-independent speaker verification
This work addresses the difficulty of implementing verification loss functions in text-independent speaker verification, offering an incremental improvement for the speech processing community.
The authors tackled the challenge of designing an effective verification loss function for deep speaker embeddings by proposing a partial AUC (pAUC) optimization approach, which achieved competitive performance with state-of-the-art identification loss functions on SITW and NIST SRE 2016 datasets.
Deep embedding based text-independent speaker verification has demonstrated superior performance to traditional methods in many challenging scenarios. Its loss functions can be generally categorized into two classes, i.e., verification and identification. The verification loss functions match the pipeline of speaker verification, but their implementations are difficult. Thus, most state-of-the-art deep embedding methods use the identification loss functions with softmax output units or their variants. In this paper, we propose a verification loss function, named the maximization of partial area under the Receiver-operating-characteristic (ROC) curve (pAUC), for deep embedding based text-independent speaker verification. We also propose a class-center based training trial construction method to improve the training efficiency, which is critical for the proposed loss function to be comparable to the identification loss in performance. Experiments on the Speaker in the Wild (SITW) and NIST SRE 2016 datasets show that the proposed pAUC loss function is highly competitive with the state-of-the-art identification loss functions.