Optimization of the area under the ROC curve using neural network supervectors for text-dependent speaker verification

Abstract

This paper explores two techniques to improve the performance of text-dependent speaker verification systems based on deep neural networks. Firstly, we propose a general alignment mechanism to keep the temporal structure of each phrase and obtain a supervector with the speaker and phrase information, since both are relevant for a text-dependent verification. As we show, it is possible to use different alignment techniques to replace the global average pooling providing significant gains in performance. Moreover, we also present a novel back-end approach to train a neural network for detection tasks by optimizing the Area Under the Curve (AUC) as an alternative to the usual triplet loss function, so the system is end-to-end, with a cost function close to our desired measure of performance. As we can see in the experimental section, this approach improves the system performance, since our triplet neural network based on an approximation of the AUC (aAUC) learns how to discriminate between pairs of examples from the same identity and pairs of different identities. The different alignment techniques to produce supervectors in addition to the new back-end approach were tested on the RSR2015-Part I database for text-dependent speaker verification, providing competitive results compared to similar size networks using the global average pooling to extract supervectors and using a simple back-end or triplet loss training.The performance in many speaker verification tasks has improved thanks to deep learning advances in signal representations [1,2] or optimization metrics [3,4,5] that have been adapted from the state-of-the-art deep learning face verification systems. In this paper, we propose alternatives to the two following aspects. First, the generation of signal representations or embeddings which preserve the subject identity and uttered phrase. Second, a metric for training the system which, as we will show, is more appropriate for a detection task.In many recent verification systems, deep neural networks (DNNs) are trained for multi-class classification. A common approach is to apply a global average reduction mechanism [1, 2, 6], which produces a vector representing the whole utterance which is called embedding. A simple back-end as a similarity metric can be applied directly for the verification process [7] or more sophisticated methods like the one presented in [8]. However, this approach does not work efficiently in text-dependent tasks since the uttered phrase is a relevant piece of information to correctly determine the identities and the system has to detect a match in the speaker and the phrase to be correct [6,9]. In our previous work [10], we have noted that part of the imprecisions may be derived from the use of the average as a representation of the utterance, and how this problem can be solved by adding a new internal layer into the deep neural network architecture which uses an alignment method to encode the temporal structure of the phrase in a supervector. In this paper, we propose a generalization ...

Language Recognition Using Triplet Neural Networks

“…Unlike the previous systems, we showed in [11] that for the recognition task, direct optimization of the AUC metric is a better and more intuitive option than traditional metrics, since this metric is a measure of the system performance. Thus, we have also used this approach for the language recognition task with the differentiable approximation of the AUC function applied in [11] to find the network parameters θ.…”

“…For training, triplet neural networks define a loss function that aims to maximize the similarity between a pair of examples belonging to the same identity or language, while minimizing the similarity with an example from another subject or language. Following the philosophy of this kind of network and applying the new loss function that we proposed in [11], we will show that the training of a system with this approach is more suitable than traditional back-ends for the language recognition task.…”

Supervector Extraction for Encoding Speaker and Phrase Information with Neural Networks for Text-Dependent Speaker Verification

“…Results confirm that the alignment as a layer within the architecture of DNN is an interesting approach since we have obtained competitive results with a straightforward and simple alignment technique that has low computational cost, so we strongly believe that can achieve better results with more powerful techniques such as GMM or DNN posteriors. As a first approximation to the proposed future work, we have an ongoing development using GMM as a new alignment technique producing good preliminary results [29].…”

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

“…The AUC optimization [17] is a special case of the pAUC optimization with α = 0 and β = 1. It is known that the performance of a speaker verification system is determined on the discriminability of the difficult trials.…”

“…The state-of-the-art text-independent speaker verification systems [1][2][3][4] use deep neural networks (DNNs) to project speech recordings with different lengths into a common low dimensional embedding space where the speakers' identities are represented. Such a method is called deep embedding, where the embedding networks have three key components-network structure [1,3,[5][6][7], pooling layer [1,[8][9][10][11][12], and loss function [13][14][15][16][17]. This paper focuses on the last part, i.e., the loss functions.…”