Optimization of the area under the ROC curve using neural network supervectors for text-dependent speaker verification. (September 2020)

• Record Type: Journal Article
• Authors: Mingote, Victoria; Miguel, Antonio; Ortega, Alfonso; Lleida, Eduardo
• Is Part Of: Computer speech & language. Volume 63(2020)
• Publication Date: 2020-09
• Subjects: Text dependent speaker verification -- Supervectors -- Alignment -- Triplet neural network -- AUC; Speech processing systems -- Periodicals; Automatic speech recognition -- Periodicals; Computers -- Periodicals; Linguistics -- Periodicals; Speech-Language Pathology -- Periodicals; Traitement automatique de la parole -- Périodiques; Reconnaissance automatique de la parole -- Périodiques; Automatic speech recognition; Speech processing systems; Electronic journals; Periodicals; 006.454
• DOI: 10.1016/j.csl.2020.101078
• Languages: English
• ISSNs: 0885-2308

Abstract

Highlights: The use of different alignment techniques inside of deep neural networks to keep the temporal structure and to encode the relevant information in a differentiable supervector. The use of Convolutional Neuronal Networks to increase the temporal context of the representation for the training process of the system. The use of a cost function close to the objective of the verification task. A set of experiments on a small database with several phrases for text-dependent phrased based speaker verification. Study of the system performance using the different alignment techniques proposed in combination with two neural network alternatives, and the novel cost function proposed contrast with the traditional triplet loss. 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 and RSR2015-Part II 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.