A5075072276- Advanced Data Compression Techniques
- Speech and Audio Processing
- Speech Recognition and Synthesis
- Advanced Wireless Communication Techniques
- Wireless Communication Networks Research
- Forest Insect Ecology and Management
- Geology and Paleoclimatology Research
- Music and Audio Processing
- Geological and Geochemical Analysis
- Entomological Studies and Ecology
- Digital Filter Design and Implementation
- Advanced Adaptive Filtering Techniques
- Insect Pheromone Research and Control
- earthquake and tectonic studies
- Maritime and Coastal Archaeology
- Botany and Plant Ecology Studies
- Lepidoptera: Biology and Taxonomy
- Electromagnetic Launch and Propulsion Technology
- Research on scale insects
- Blind Source Separation Techniques
- Risk and Safety Analysis
- Insect Pest Control Strategies
- African Botany and Ecology Studies
- Plant Physiology and Cultivation Studies
- Dermatological and Skeletal Disorders
Université de Pau et des Pays de l'Adour
2023-2024
Laboratoire des Fluides Complexes et leurs Réservoirs
2024
Laboratoire de Dynamique des Fluides
2024
Institut National de Recherche Forestière
2015-2020
Institut National de Recherche pour l'Agriculture, l'Alimentation et l'Environnement
2020
Hôpital Européen
2013
University of California, Los Angeles
1999-2008
Texas Instruments (United States)
2004
Abstract Large explosive volcanic eruptions from island arcs pour pyroclastic currents into marine basins, impacting ecosystems and generating tsunamis that threaten coastal communities infrastructures. Risk assessments require robust records of such highly hazardous events, which is challenging as most the products lie buried under sea. Here we report discovery by IODP Expedition 398 a giant rhyolitic pumice deposit emplaced 520 ± 10 ky ago at water depths 200 to 1000 m during...
Abstract Caldera-forming eruptions of silicic volcanic systems are among the most devastating events on Earth. By contrast, post-collapse activity initiating new caldera cycles is generally considered less hazardous. Formed after Santorini’s latest caldera-forming eruption ~1600 bce , Kameni Volcano in southern Aegean Sea enables eruptive evolution a recharging multi-cyclic to be reconstructed. Kameni’s record has been documented by onshore products and historical descriptions mainly...
We present a framework for developing source coding, channel coding and decoding as well erasure concealment techniques adapted distributed (wireless or packet-based) speech recognition. It is shown that recognition opposed to more sensitive errors than erasures, appropriate design criteria are determined. For decoding, we introduce novel technique combining at the receiver soft decision with error detection. Frame used decoder deal unreliable frames. At stage, modify engine itself take into...
Santorini Island (Greece) is an active volcano which has alternated between dormant and periods over the last 650,000 years with latest volcanic unrest occurring in 2011–2012. Here we report a geochemical survey of fumarolic gases collected at Nea Kameni islet located center caldera period 2015–2022 order to study activity changes hydrothermal conditions. This marked by absence significant anomalies compared 2011–2012, implying that no new magma upwelling occurred. evident from low CO 2 /CH...
We introduce the concept of joint channel decoding and Viterbi recognition, by which recognizer is modified to take into account confidence in decoded feature after transmission. present a metric for evaluating such based on soft decision decoding. As case study, we quantize MFCCs using predictive VQ. The overall sourcechannel coding scheme operating at combined rate 1 kbps shown provide good recognition accuracy over wide range Rayleigh fading channels.
This paper presents source and channel coding techniques for remote automatic speech recognition (ASR) systems. As a case study, line spectral pairs (LSP) extracted from the 6th order all-pole perceptual linear prediction (PLP) spectrum are transmitted features then obtained. The LSPs, quantized using first-order predictive vector quantization (VQ) at 300 bps, provide accuracy comparable to that of baseline system with no quantization. A new soft decision decoding scheme appropriate is...
This paper presents a multi-rate wideband speech coder with bit rates from 8 to 32 kb/s. The uses splitband approach, where the input signal, sampled at 16 kHz, is split into two equal frequency bands 0-4 kHz and 4-8 each of which decimated an sampling rate. lower band coded using adaptive (AMR) family high-quality narrowband coders, while higher represented by simple but effective parametric model. A complete solution including this coder, channel coding for various GSM channels, dynamic...
This paper presents bandwidth-efficient speech transmission systems using rate-compatible channel coders and variable bitrate embedded source coders. Rate-compatible punctured convolutional codes (RCPC) are often used to provide unequal error protection (UEP) via progressive bit puncturing. RCPC well suited for constellations which Euclidean Hamming distances equivalent (BPSK 4-PSK). introduces trellis (RCPT) where rate compatibility UEP provided puncturing of symbols in a trellis. RCPT...
The performance of speech recognition systems trained in quiet degrades signicantly under noisy conditions. To address this problem, a Weighted Viterbi Recognition (WVR) algorithm that is function the SNR each frame proposed. Acoustic models on clean data, and acoustic front-end features are kept unchanged approach. Instead, condence/rob ustness factor assigned to output observation probability according its estimate during decoding stage. Comparative experiments conducted with different...
We present a back-end solution developed at Texas Instruments for noise robust speech recognition. The consists of three techniques: 1) joint additive and convolutive compensation (JAC) which adapts acoustic models; 2) an enhanced channel estimation procedure extends JAC performance towards lower SNR ranges; 3) N-pass decoding algorithm. the proposed is evaluated on Aurora-2 database. With 20% fewer model parameters without need second order derivative recognition features, 91.86%,...
This paper presents an embedded joint source-channel coding scheme of speech. The source coder is variable bit rate perceptually based sub-band producing bits with different error sensitivities. channel encoder a compatible punctured trellis code (RCPT) which permits variability and unequal protection by puncturing symbols. Furthermore, RCPT design naturally incorporates large constellations, allowing high information per symbol. speech the symbols provide embeddibility scheme. robust to...
We present a novel joint source-channel coding scheme for speech signals. The source coder is perceptually based sub-band producing bits with different error sensitivities. channel encoder rate compatible punctured trellis code (RCPT) that permits unequal protection. RCPT design naturally incorporates large constellations, allowing high information per symbol. robust to acoustic noise, adapts automatically every 20 ms and produces good quality wide range of conditions. paper presents method...
Training triphone acoustic models for speech recognition is time-consuming and requires important manual intervention. We present an alternative solution, performing automatic training by use of a pronunciation phonetic table which summarizes the articulatory characteristics target language. The method able to train triphones any language, given existing set reference monophones in one or more languages, automatically tasks monophone seeding, clustering other steps. nature algorithm lends...
Helium-rich gases (up to 6%) upwell along a fault that bounds the western edge of granitic Morvan Massif (northeastern part French Central) and sedimentary rocks southeastern Paris Basin. The sampled gas is mainly composed nitrogen (90 %.). radiogenic 3He/4He (0.02 Ra, where Ra atmospheric ratio) nucleogenic 21Ne/22Ne (0.031) imply basement-derived, in good agreement with presence granite area. high He concentration can be explained by rock/ water system isolated over geological time scales...
This paper presents a novel multi-band CELP coder with the following characteristics: wideband coding (6.5 kHz), variable bit rate (VBR) (10-24 kbps), low-delay (10 ms), embeddibility, and perceptually based dynamic allocation. The excitation signal of linear prediction filter is vector sum eight off-line pre-filtered bandpass vectors. codebooks are tree structured, providing embeddibility rate. allocation bitstream among different bands on perceptual importance each band. structure scheme...