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

In this paper, we explore self-supervised learning (SSL) for analyzing a first-of-its-kind database of cry recordings containing clinical indications of more than a thousand newborns. Specifically, we target cry-based detection of neurological injury as well as identification of cry triggers such as pain, hunger, and discomfort.

Artificial Intelligent (AI) tools can be useful to address model deficits in the design of communication systems. However, conventional learning-based AI algorithms yield poorly calibrated decisions, unabling to quantify their outputs uncertainty. While Bayesian learning can enhance calibration by capturing epistemic uncertainty caused by limited data availability, formal calibration guarantees only hold under strong assumptions about the ground-truth, unknown, data generation mechanism.

Even without a separate tracking algorithm, the directions of arrival (DOAs) of moving talkers can be estimated with a deep neural network (DNN) when the movement trajectories used for training allow the generalization to real signals. Previously, we proposed a framework for generating training data with time-variant source activity and sudden DOA changes. Slowly moving sources could be seen as a special case thereof, but were not explicitly modeled. In this paper, we extend this framework by using small jumps between neighboring discrete DOAs to simulate gradual movements.

False data injection attacks (FDIAs) on smart power grids' measurement data present a threat to system stability. When malicious entities launch cyberattacks to manipulate the measurement data, different grid components will be affected, which leads to failures. For effective attack mitigation, two tasks are required: determining the status of the system (normal operation/under attack) and localizing the attacked bus/power substation. Existing mitigation techniques carry out these tasks separately and offer limited detection performance.

Photographs taken in low-illumination environment have a low signal-to-noise ratio and impaired visual quality. Enhancing low-light images tends to amplify noise. To address this problem, we propose a Multi-Scale Recursive Feature Enhancement (MSRFE) network for low light image enhancement. The MSRFE network consists of several Feature Enhancement (FE) blocks which are applied to enhance the multi-scale image feature and remove the noise recursively in each scale residual map between adjacent scale feature.

We propose a novel approach for blind room impulse response (RIR) estimation systems in the context of a downstream application scenario, far-field automatic speech recognition (ASR). We first draw the connection between improved RIR estimation and improved ASR performance, as a means of evaluating neural RIR estimators.

Recently, many studies have been conducted on automated epileptic seizures detection. However, few of these techniques are applied in clinical settings for several reasons. One of them is the imbalanced nature of the seizure detection task. Additionally, the current detection techniques do not really generalize to other patient populations. To address these issues, we present in this paper a hybrid CNN-LSTM model robust to cross-site variability. We investigate the use of data augmentation (DA) methods as an efficient tool to solve imbalanced training problems.

Recently, many studies have been conducted on automated epileptic seizures detection. However, few of these techniques are applied in clinical settings for several reasons. One of them is the imbalanced nature of the seizure detection task. Additionally, the current detection techniques do not really generalize to other patient populations. To address these issues, we present in this paper a hybrid CNN-LSTM model robust to cross-site variability. We investigate the use of data augmentation (DA) methods as an efficient tool to solve imbalanced training problems.

This paper studies distributed online learning under Byzantine attacks. Performance of an online learning algorithm is characterized by (adversarial) regret, and a sublinear bound is preferred. But we prove that, even with a class of state-of-the-art robust aggregation rules, in an adversarial environment and with Byzantine participants, distributed online gradient descent can only achieve a linear adversarial regret bound, which is tight. This is the inevitable consequence

Vector-quantized autoencoders have recently gained interest in image compression, generation and self-supervised learning. However, as a neural compression method, they lack the possibility to allocate a variable number of bits to each image location, e.g. according to the semantic content or local saliency. In this paper, we address this limitation in a simple yet effective way. We adopt a product quantizer (PQ) that produces a set of discrete codes for each image patch rather than a single index.