Advanced Signal Processing for Biomedical Engineering, Neurosciences, Geosciences, and Communication Systems
The Signal research team is composed of 6 permanent members and many PhD students. The idiosyncrasy of our research is to propose a collaborative framework with other laboratories in our application fields to develop advanced, innovative and adapted tools for the processing of signals or images acquired with biomedical sensor networks (cardiology, neurosciences) or in geosciences (seismology and marine ecology), but also in wireless communications. The approach developed relies on various signal processing techniques such as frequency, time-frequency or time-scale analysis, modulation spectrum, delay time estimation, single/multidimensional signal modeling, generalized linear regression approaches, or tensor-based methods.
Our main application domains are with cardiology (electrocardiogram, electrogram), nervous (neurogram), muscular (electromyogram) and respiratory electrical activity, and for the study of cerebral/cerebellar function by electroencephalography (EEG), but also using functional near-infrared spectrometry (fNIRS) and functional magnetic resonance imaging (fMRI). Our specialization relates to multi-sensor methods, tensor decompositions, principal component / independent analysis, with elements of interpretability / explicability, robust methods (e.g., L1-based), the joint processing of multimodal data, notably in the context of functional acquisitions of cerebellar activity (EEG, fNIRS, fMRI), and of invasive (intracardiac electrograms) and non-invasive (surface electrocardiogram) modalities of cardiac activity for the characterization of arrhythmias.
In the field of neurosciences, our current main interest lies around the study of functional cerebellar asymmetries in close collaboration with the Service d'exploration fonctionnelle du système nerveux (EFSN) of the CHU Nice Pasteur within the framework of the MuExC3Po project (H2020 EU Cofund 2020-2023). We were the first to prove the feasability of recording cerebellar activity using fNIRS, and now aim to perform the first multimodal (EEG/fNRIS/fMRI) acquisitions of cerebellar neural activity. Signal is also involved in the national project ANR NeuroInfPredict (2022-2026) with the IPMC lab (UCA/CNRS/Inserm) around the analysis of electrical activity (neurogram) recorded on the splenic nerve in order to detect changes in the characteristics of nerve depolarizations. This change may be caused by the detection, through the efferent nerves of an inflammatory phenomenon (typ. rheumatoid arthritis). The processing techniques considered are based on filtering, pattern analysis, classification and supervised and unsupervised learning.
In cardiology, our research interest lays on supraventricular arrhythmias (atrial fibrillation and atrial flutter), which constitute a major societal challenge due to their growing incidence and economic impact. In partnership with CHU Nice Pasteur and CHPG of Monaco, signal processing and machine learning tools are currently being developed for the characterization of these cardiac disorders, to assist in diagnosis and interventional procedures in catheter ablation therapy. These tools are at the heart of an on-going project with some Malaysian Universities, to support the growth of e-health in the Southeast Asian region. An industrial collaboration is starting around our algorithms for automatic classification of atrial electrograms for the ablation of persistent atrial fibrillation,towards an integration in a commercial software library proposed with some new device for the acquisition of very low noise electrophysiological signals. These themes are also linked with the 3IA chair of one member.
In collaboration with GeoAzur and Ecoseas UCA labs, we developed adapted tools for automatic analysis/detection/classification of multiple seismic and marine phenomena recorded by seismometer/hydrophone networks (earthquakes, seismic swarms, tremors, but also marine mammal vocalizations, anthropogenic noises such as ship engines, vibrations from coastal airports, . .., as well as underwater and surface meteorology). A first study focused on the consequences of anthropogenic noise in the Mediterranean Sea on both the coastal species studied by EcoSeas, and on the large marine mammals of the Pelagos Sanctuary (especially fin whale vocalizations in the Mediterranean Sea) from the very large amounts of seismic and underwater acoustic data (> 100~TB annually) collected by GeoAzur during its underwater seismic exploration campaigns. Our team is also involved in national project ANR Fluid2Slip (2021-2025) with GeoAzur and Ifremer to develop new advanced tools to study the fluid circulation in subduction zones and submarine seismic faults. The objective of the project is to study and derive relationships between seismicity, transient signals and physical properties of faults such as fluid content and/or fluid circulation. Signal is involved in the development of new algorithmic schemes to automatically detect, extract, classify and localize the different seismic signals recorded by a dense network of seafloor seismometers and land stations in Ecuador: earthquakes, transient signals such as non-volcanic tremors, low-frequency (LFE) and very low-frequency (VLFE) tremors, as well as seismic repeaters.
Tensors for Signal Processing
Another transversal research activity of our team involves the development of new tensor models with applications in various domains (identification of both linear and non-linear systems, biomedical signal processing, road traffic prediction and design of cooperative MIMO communication systems, typically 5G/6G). The theoretical perspectives of this research include the study of tensor models of reduced complexity in the form of tensor networks, and low rank tensor approximation methods; the development of tensor completion methods for the reconstruction of missing data in high-dimensional databases; a deepening of tensor approaches for the modeling of nonlinear and polynomial systems; applications of tensors to the design of 6th generation (6G) MIMO communication systems. This research is conducted in collaboration with some other teams: Cristal lab. (University of Lille), GIPSA-Lab (Univ. Grenoble). Internationally, a network has been developed within the framework of the CAPES-COFECUB program between France and Brazil, between UCA and the Federal University of Ceara (UFC), the University of Brasilia (UnB) and the State University of Campinas (UNICAMP).