Triplet Loss Based Cosine Similarity Metric Learning for Text-independent Speaker Recognition

Abstract

Deep neural network based speaker embeddings become increasingly popular in the text-independent speaker recognition task. In contrast to a generatively trained i-vector extractor, a DNN speaker embedding extractor is usually trained discriminatively in the closed set classification scenario using softmax. The problem we addressed in the paper is choosing a dnn based speaker embedding backend solution for the speaker verification scoring. There are several options to perform speaker verification in the dnn embedding space. One of them is using a simple heuristic speaker similarity metric for scoring (e.g. cosine metric). Similarly with i-vector based systems, the standard Linear Discriminant Analisys (LDA) followed by the Probabilistic Linear Discriminant Analisys (PLDA) can be used for segregating speaker information. As an alternative, the discriminative metric learning approach can be considered. This work demonstrates that performance of deep speaker embeddings based systems can be improved by using Cosine Similarity Metric Learning (CSML) with the triplet loss training scheme. Results obtained on Speakers in the Wild and NIST SRE 2016 evaluation sets demonstrate superiority and robustness of CSML based systems.

Partial AUC Optimization Based Deep Speaker Embeddings with Class-Center Learning for Text-Independent Speaker Verification

“…Results demonstrated that the proposed pAUC deep embedding is highly competitive in performance with the state-of-the-art identification loss based deep embedding methods with the Softmax and ArcSoftmax output units. Note that a very recent work proposed at the same time as our work in [23] maximizes the area under the ROC curve (AUC) for text-dependent speaker verification [20]. It can be shown that AUC is a particular case of pAUC and experimental results show the pAUC deep embedding outperforms the AUC deep embedding significantly.…”

A Discriminative Condition-Aware Backend for Speaker Verification

“…where s k is the score for trial k given by Equation (2), σ is the sigmoid function, and α and β are the calibration parameters, trained to minimize the quantity in Equation (3). To summarize, Equations (1), (2) and (5) show the pipeline that is applied to the embeddings in the standard PLDA-based backend. The parameters involved in these equations are all trained separately, freezing the parameters of the previous steps in order to obtain input data to train the next step.…”

“…We propose a backend with the same functional form as the PLDAbackend explained in the previous section, but where all parameters are optimized jointly, in a manner similar to the one used in [6] (though, note that in this paper we only optimize jointly up to the backend stage instead of the full pipeline, as in Rohdin's paper). We first initialize all parameters in Equations (1), (2) and (5) as in the standard PLDA-based backend. Then, we fine tune the parameters to optimize the cross-entropy in Equation (3) using some variant of stochastic gradient descent.…”

“…This problem has been widely acknowledged in the community and several papers have been published with different attempts to eliminate or integrate some of these stages. Angular and triplet losses [2,3,4,5] have been used to train the speaker embedding extractor DNN instead of the standard cross-entropy with the goal of making the backend stages unnecessary. The cosine distance between embeddings resulting from these losses can be directly used to generate a score for each trial without the need for a separate backend.…”

Coarse-to-Fine Optimization for Speech Enhancement

“…It has quite a few applications including noise cancelling, audio editing, preprocessing for speech recognition, just to name a few. Denote the noisy speech as y(t), we have y(t) = x(t) + n(t) (1) where x(t) and n(t) are respectively the clean speech and the noise, with t being the time index. Speech enhancement tries to recover the clean speech x 1 from the noisy speech y.…”