Audio and visual modalities are inherently connected in speech signals: lip movements and facial expressions are correlated with speech sounds. This motivates studies that incorporate the visual modality to enhance an acoustic speech signal or even restore missing audio information. Specifically, this paper focuses on the problem of audio-visual speech inpainting, which is the task of synthesizing the speech in a corrupted audio segment in a way that it is consistent with the corresponding visual content and the uncorrupted audio context. We present an audio-visual transformer-based deep learning model that leverages visual cues that provide information about the content of the corrupted audio. It outperforms the previous state-of-the-art audio-visual model and audio-only baselines. We also show how visual features extracted with AV-HuBERT, a large audiovisual transformer for speech recognition, are suitable for synthesizing speech.

Comunicació presentada a Interspeech 2022, celebrat del 18 al 22 de setembre de 2022 a Inchon, Corea del Sud.

A fairly straightforward approach for music source separation is to train independent models, wherein each model is dedicated for estimating only a specific source. Training a single model to estimate multiple sources generally does not perform as well as the independent dedicated models. However, Conditioned U-Net (C-U-Net) uses a control mechanism to train a single model for multi-source separation and attempts to achieve a performance comparable to that of the dedicated models. We propose a multi-channel U-Net (M-U-Net) trained using a weighted multi-task loss as an alternative to the C-U-Net. We investigate two weighting strategies for our multi-task loss: 1) Dynamic Weighted Average (DWA), and 2) Energy Based Weighting (EBW). DWA determines the weights by tracking the rate of change of loss of each task during training. EBW aims to neutralize the effect of the training bias arising from the difference in energy levels of each of the sources in a mixture. Our methods provide three-fold advantages compared to C-U-Net: 1) Fewer effective training iterations per epoch, 2) Fewer trainable network parameters (no control parameters), and 3) Faster processing at inference. Our methods achieve performance comparable to that of C-U-Net and the dedicated U-Nets at a much lower training cost.

In this paper, we present a new dataset of music performance videos which can be used for training machine learning methods for multiple tasks such as audio-visual blind source separation and localization, cross-modal correspondences, cross-modal generation and, in general, any audio-visual self-supervised task. These videos, gathered from YouTube, consist of solo musical performances of 13 different instruments. Compared to previously proposed audio-visual datasets, Solos is cleaner since a big amount of its recordings are auditions and manually checked recordings, ensuring there is no background noise nor effects added in the video post-processing. Besides, it is, up to the best of our knowledge, the only dataset that contains the whole set of instruments present in the URMP dataset, a high-quality dataset of 44 audio-visual recordings of multi-instrument classical music pieces with individual audio tracks. URMP was intented to be used for source separation, thus, we evaluate the performance on the URMP dataset of two different source-separation models trained on Solos. The dataset is publicly available at https://juanfmontesinos.github.io/Solos/.