A5082784886- Speech and Audio Processing
- Blind Source Separation Techniques
- Speech Recognition and Synthesis
- Music and Audio Processing
- Advanced Adaptive Filtering Techniques
- Stellar, planetary, and galactic studies
- S100 Proteins and Annexins
- Medical History and Innovations
- Retinopathy of Prematurity Studies
- Burn Injury Management and Outcomes
- Astronomy and Astrophysical Research
- Immune Response and Inflammation
- History of Medicine Studies
- Data Management and Algorithms
- Historical and Scientific Studies
- Sepsis Diagnosis and Treatment
- Connective tissue disorders research
- Astronomical Observations and Instrumentation
- Metaheuristic Optimization Algorithms Research
- Astrophysics and Star Formation Studies
- Constraint Satisfaction and Optimization
- Trauma and Emergency Care Studies
- Cerebral Venous Sinus Thrombosis
- Gut microbiota and health
- Philosophy, Sociology, Political Theory
Télécom Paris
2019-2025
Laboratoire Traitement et Communication de l’Information
2022-2025
Centre Hospitalier Saint-Joseph Saint-Luc
2014-2024
Signal Processing (United States)
2024
RIKEN Center for Advanced Intelligence Project
2020-2024
Erytech Pharma (France)
2024
Statistics Belgium
2022-2024
Questel (France)
2023
Neurophysiologie respiratoire expérimentale et clinique
2022
Assistance Publique – Hôpitaux de Paris
2022
This study investigates the influence of Herbig-Haro jets on initiating star formation in dense environments. When molecular clouds are nearing gravitational instability, impact a protostellar jet could provide impetus needed to catalyse formation. A high-energy-density experiment was carried out at LULI2000 laser facility, where supersonic generated by nanosecond used compress foam or plastic ball, mimicking interaction with cloud. Simulations using 3D radiation hydrodynamics code TROLL...
This paper introduces a new training strategy to improve speech dereverberation systems using minimal acoustic information and reverberant (wet) speech. Most existing algorithms rely on paired dry/wet data, which is difficult obtain, or target metrics that may not adequately capture reverberation characteristics can lead poor results non-target metrics. Our approach uses limited information, like the time (RT60), train system. The system's output resynthesized generated room impulse response...
We present an analysis of the Far Ultraviolet Spectroscopic Explorer (FUSE) spectra five V1093 Her (PG 1716+426) stars, subgroup hydrogen-rich subdwarf B stars that exhibit very low amplitude, long-period luminosity variations. Our primary aim is to investigate whether these display abundances which differ from those observed in nonvariable sdB and also shorter period V361 Hya variables. For light elements for beyond iron peak, our are consistent with trends earlier studies. important peak...
To provide recommendations to facilitate the management of severe thermal burns during acute phase in adults and children. A committee 20 experts was asked produce six fields burn management, namely, (1) assessment, admission specialised centres, telemedicine; (2) haemodynamic management; (3) airway smoke inhalation; (4) anaesthesia analgesia; (5) wound treatments; (6) other treatments. At start recommendation-formulation process, a formal conflict-of-interest policy developed enforced...
Clone detection is usually applied in the context of detecting small-to medium scale fragments duplicated code large software systems. In this paper, we address problem clone to plagiarism source assignments done by computer science students. Plagiarism comes with a distinct set constraints usual approaches, which influenced design approach present paper. For instance, can be heavily changed at superficial level (in an attempt look genuine), yet functionally very similar. Since turned...
Septic syndromes are the leading cause of death in intensive care units. They characterized by development immune dysfunctions such as endotoxin tolerance (ET), whose intensity and duration associated with increased risk nosocomial infections mortality. Alarmins S100A8 S100A9 have been shown to be after septic shock. Importantly, a delayed mRNA increase predicts hospital-acquired infection patients. The aim this study was investigate regulation expression an ex vivo model ET.ET reproduced...
This paper describes aneural blind source separation (BSS) method based on amortized variational inference (AVI) of a non-linear generative model mixture signals. A classical statistical approach to BSS is fit linear that consists spatial and models representing the inter-channel covariances power spectral densities sources, respectively. Although autoencoder (VAE) has successfully been used as with latent features, it should be pretrained from sufficient amount isolated Our method, in...
Septic shock remains a serious disease with high mortality and increased risk of hospital-acquired infection. The prediction outcome is the utmost importance for selecting patients therapeutic strategies aiming to modify immune response. aim this study was assess capability S100A9 messenger RNA in whole blood from septic predict survival occurrence infection.Cohort study.Two intensive care units university hospital.The included (n = 166) healthy volunteers 44).None.For patients, overall 38%...
This letter describes a time-varying extension of independent vector analysis (IVA) based on the normalizing flow (NF), called NF-IVA, for determined blind source separation multichannel audio signals. As in IVA, NF-IVA estimates demixing matrices that transform mixture spectra to complex-valued spatial domain such likelihood those is maximized under some non-Gaussian model. While IVA performs time-invariant bijective linear transformation, series transformations (flow blocks) adaptively...
We present the results of an observational campaign for long-period variable subdwarf B star PG 1338+481. Seven continuous weeks observing time at Steward Observatory 1.55 m Kuiper telescope on Mount Bigelow, Arizona, and 1.3 MDM Kitt Peak rendered ~250 hr simultaneous U/R series photometry, as well extra ~70 R-band-only data. The analysis combined light curves resulted in extraction 13 convincing periodicities 2100-7200 s range, with amplitudes up to ~0.3% ~0.2% U R, respectively. Comparing...
This paper describes a joint blind source separation and dereverberation method that works adaptively efficiently in reverberant noisy environment. The modern approach to (BSS) is formulate probabilistic model of multichannel mixture signals consists representing the time-frequency structures spectrograms spatial inter-channel covariance images. cutting-edge BSS this thread research fast multi-channel nonnegative matrix factorization (FastMNMF) low-rank based on (NMF) full-rank...
This paper describes a method for estimating the room impulse response (RIR) microphone and sound source located at arbitrary positions from 3D mesh data of room. Simulating realistic RIRs with pure physics-driven methods often fails balance between physical consistency computational efficiency, hindering application to real time speech processing. Alternatively, one can use MESH2IR, fast black-box estimator that consists an encoder extracting latent code graph convolutional network (GCN)...
Single-channel speech dereverberation aims at extracting a dry signal from recording affected by the acoustic reflections in room. However, most current deep learning-based approaches for are not interpretable room acoustics, and can be considered as black-box systems that regard. In this work, we address problem regularizing training loss using novel physical coherence which encourages impulse response (RIR) induced dereverberated output of model to match properties was recorded. Our...
Tree decomposition introduced by Robertson and Seymour aims to decompose a problem into clusters constituting an acyclic graph. There are works exploiting tree for complete search methods. In this paper, we show how can be used efficiently guide local methods that use large neighborhoods like VNS. We propose DGVNS (Decomposition Guided VNS) which uses the graph of in order build neighborhood structures enabling better diversification intensification. Second, introduce tightness dependent...
Diffusion models are receiving a growing interest for variety of signal generation tasks such as speech or music synthesis. WaveGrad, example, is successful diffusion model that conditionally uses the mel spectrogram to guide process high-fidelity audio. However, face important challenges concerning noise training and inference, they have difficulty generating high-quality speakers were not seen during training. With aim minimizing conditioning error increasing efficiency process, we propose...