Sorry, you need to enable JavaScript to visit this website.

Recently, Self-Supervised Features (SSF) trained on extensive speech datasets have shown significant performance gains across various speech processing tasks. Nevertheless, their effectiveness in Speech Enhancement (SE) systems is often suboptimal due to insufficient optimization for noisy environments. To address this issue, we present a novel methodology that directly utilizes SSFs extracted from clean speech for enhancing SE models. Specifically, we leverage the clean SSFs for latent space modeling within the Conditional Variational Autoencoder (CVAE) framework.

Filter Decomposition (FD) methods have gained traction in compressing large neural networks by dividing weights into basis and coefficients. Recent advancements have focused on reducing weight redundancy by sharing either basis or coefficients stage-wise. However, traditional sharing approaches have overlooked the potential of sharing basis on a network-wide scale. In this study, we introduce an FD technique called G-SharP that elevates performance by using globally shared kernels throughout the network.

This paper describes SRCB-LUL team's unsupervised relapse detection system submitted to the 2nd E-Prevention Challenge (Psychotic and Non-Psychotic Relapse Detection using Wearable-Based Digital Phenotyping). In our system, a person identification task is added to make the feature extraction network better distinguish between different behavior patterns. Three different structures of the feature extraction network are adopted. Then, the extracted features are used to train an Elliptic Envelope model of each patient for anomaly detection.

Speech enhancement systems are typically trained using pairs of clean and noisy speech. In audio-visual speech enhancement

The use of AI has led to the era of pervasive intelligence, marked by a proliferation of smart devices in our daily lives. Federated Learning (FL) enables machine learning at the edge without having to share user-specific private data with an untrusted third party. Conventional FL techniques are supervised learning methods, where a fundamental challenge is to ensure that data is reliably annotated at the edge. Another approach is to obtain rich and informative representations ofunlabeled data, which is suitable for downstream tasks.

Federated learning (FL) allows different participants to collaborate on model training without transmitting raw data, thereby protecting user data privacy. However, FL faces a series of security and privacy issues (e.g. the leakage of raw data from publicly shared parameters). Several privacy protection technologies, such as homomorphic encryption, differential privacy and functional encryption, are introduced for privacy enhancement in FL. Among them, the FL frameworks based on functional encryption better balance security and performance, thus receiving increasing attention.

The correlation between the sharpness of loss minima and generalisation in the context of deep neural networks has been subject to discussion for a long time. Whilst mostly investigated in the context of selected benchmark data sets in the area of computer vision, we explore this aspect for the acoustic scene classification task of the DCASE2020 challenge data. Our analysis is based on two-dimensional filter-normalised visualisations and a derived sharpness measure.

Convolutional neural networks (CNNs) have long been the paradigm of choice for robust medical image processing (MIP). Therefore, it is crucial to effectively and efficiently deploy CNNs on devices with different computing capabil- ities to support computer-aided diagnosis. Many methods employ factorized convolutional layers to alleviate the bur- den of limited computational resources at the expense of expressiveness.

We describe our contribution to the 2nd e-Prevention challenge, which focuses on the unsupervised non-psychotic (Track 1) and psychotic (Track 2) relapse detection using wearable-based digital phenotyping. We exploit the measurements gathered from the gyroscope, the accelerometer, and the heart rate-related sensors embedded in a smartwatch. We also include the available sleep information in our experiments. Four dedicated autoencoders are trained to learn embedded representations from each one of the considered modalities.

Crafting an effective Automatic Speech Recognition (ASR) solution for dialects demands innovative approaches that not only address the data scarcity issue but also navigate the intricacies of linguistic diversity. In this paper, we address the aforementioned ASR challenge, focusing on the Tunisian dialect. First, textual and audio data is collected and in some cases annotated.