This paper presents a new approach based on recurrent neural networks (RNN) to the multiclass audio segmentation task whose goal is to classify an audio signal as speech, music, noise or a combination of these. The proposed system is based on the use of bidirectional long short-term Memory (BLSTM) networks to model temporal dependencies in the signal. The RNN is complemented by a resegmentation module, gaining long term stability by means of the tied state concept in hidden Markov models. We explore different neural architectures introducing temporal pooling layers to reduce the neural network output sampling rate. Our findings show that removing redundant temporal information is beneficial for the segmentation system showing a relative improvement close to 5%. Furthermore, this solution does not increase the number of parameters of the model and reduces the number of operations per second, allowing our system to achieve a real-time factor below 0.04 if running on CPU and below 0.03 if running on GPU. This new architecture combined with a data-agnostic data augmentation technique called mixup allows our system to achieve competitive results in both the Albayzín 2010 and 2012 evaluation datasets, presenting a relative improvement of 19.72% and 5.35% compared to the best results found in the literature for these databases.

Area under the ROC curve (AUC) optimisation techniques developed for neural networks have recently demonstrated their capabilities in different audio and speech related tasks. However, due to its intrinsic nature, AUC optimisation has focused only on binary tasks so far. In this paper, we introduce an extension to the AUC optimisation framework so that it can be easily applied to an arbitrary number of classes, aiming to overcome the issues derived from training data limitations in deep learning solutions. Building upon the multiclass definitions of the AUC metric found in the literature, we define two new training objectives using a one-versus-one and a one-versus-rest approach. In order to demonstrate its potential, we apply them in an audio segmentation task with limited training data that aims to differentiate 3 classes: foreground music, background music and no music. Experimental results show that our proposal can improve the performance of audio segmentation systems significantly compared to traditional training criteria such as cross entropy.

This paper proposes a direct text to speech translation system using discrete acoustic units. This framework employs text in different source languages as input to generate speech in the target language without the need for text transcriptions in this language. Motivated by the success of acoustic units in previous works for direct speech to speech translation systems, we use the same pipeline to extract the acoustic units using a speech encoder combined with a clustering algorithm. Once units are obtained, an encoder-decoder architecture is trained to predict them. Then a vocoder generates speech from units. Our approach for direct text to speech translation was tested on the new CVSS corpus with two different text mBART models employed as initialisation. The systems presented report competitive performance for most of the language pairs evaluated. Besides, results show a remarkable improvement when initialising our proposed architecture with a model pre-trained with more languages.

Speech Activity Detection (SAD) aims to accurately classify audio fragments containing human speech. Current state-of-the-art systems for the SAD task are mainly based on deep learning solutions. These applications usually show a significant drop in performance when test data are different from training data due to the domain shift observed. Furthermore, machine learning algorithms require large amounts of labelled data, which may be hard to obtain in real applications. Considering both ideas, in this paper we evaluate three unsupervised domain adaptation techniques applied to the SAD task. A baseline system is trained on a combination of data from different domains and then adapted to a new unseen domain, namely, data from Apollo space missions coming from the Fearless Steps Challenge. Experimental results demonstrate that domain adaptation techniques seeking to minimise the statistical distribution shift provide the most promising results. In particular, Deep CORAL method reports a 13% relative improvement in the original evaluation metric when compared to the unadapted baseline model. Further experiments show that the cascaded application of Deep CORAL and pseudo-labelling techniques can improve even more the results, yielding a significant 24% relative improvement in the evaluation metric when compared to the baseline system.

This paper describes a post-evaluation analysis of the system developed by ViVoLAB research group for the IberSPEECH-RTVE 2020 Multimodal Diarization (MD) Challenge. This challenge focuses on the study of multimodal systems for the diarization of audiovisual files and the assignment of an identity to each segment where a person is detected. In this work, we implemented two different subsystems to address this task using the audio and the video from audiovisual files separately. To develop our subsystems, we used the state-of-the-art speaker and face verification embeddings extracted from publicly available deep neural networks (DNN). Different clustering techniques were also employed in combination with the tracking and identity assignment process. Furthermore, we included a novel back-end approach in the face verification subsystem to train an enrollment model for each identity, which we have previously shown to improve the results compared to the average of the enrollment data. Using this approach, we trained a learnable vector to represent each enrollment character. The loss function employed to train this vector was an approximated version of the detection cost function (aDCF) which is inspired by the DCF widely used metric to measure performance in verification tasks. In this paper, we also focused on exploring and analyzing the effect of training this vector with several configurations of this objective loss function. This analysis allows us to assess the impact of the configuration parameters of the loss in the amount and type of errors produced by the system.