Référence bibliographique :

M. Vacher, F. Portet, A. Fleury, N. Noury, Challenges in the Processing of Audio Channels for Ambient Assisted Living, HealthCom'10, 1-2 July 2010, Lyon, France, pp. 330-337.

Abstract - One of the greatest challenges in Ambient Assisted Living is to design health smart homes that could be able to anticipate the needs of its inhabitant while maintaining their comfort and their safety with an adaptation of the house environment and a facilitation of the connections to the outside world. The most likely to benefit from these smart homes are people in loss of autonomy such as the disabled people or the elderly with cognitive deficiencies. But it becomes essential to ease the interactions with the smart home through dedicated interfaces, in particular, thanks to systems reactive to vocal orders. Audio recognition is also a promising way to ensure more safety by contributing to detection of distress situations. This paper presents the stakes and the challenges of this domain based on some experiments carried out concerning distress call recognition and sound classification at home.

Référence bibliographique :

A. Fleury, N. Noury, M. Vacher, Introducing Knowledge in the Process of Supervised Classification of Activities of Daily Living in Health Smart Homes , HealthCom'10, 1-2 July 2010, Lyon, France, pp. 322-329.

Abstract - Telemedicine and Telemonitoring of elderly people is an actual challenge that is explored to prevent some problems linked to the constant growing of the mean age of the population. It requires to recognize the behavior and the actions of a person inside his own home with non-intrusive sensors and to process data to check the evolution of the person. Activities of Daily Living can be learned and automatically recognized using supervised classification on sensor data. This paper presents the results of the study of prior introduction, in Support Vector Machine, to improve this automatic recognition of Activities of Daily Living. We started from a set of data acquired in daily life during an experimentation in the Health Smart Home of the TIMC-IMAG Lab. From this restricted set of data, we obtained models for seven activities of Daily Living and test, with leave- one-out method, the performance of this classification. This first step gave baseline results that this paper tends to improve using consistent priors to compute more specific and accurate models of the different activities that are learned and obtain better results on the leave-one-out method on the sensors data.

Référence bibliographique :

A. Fleury, M. Vacher, F. Portet, P. Chahuara, N. Noury, A Multimodal Corpus Recorded in a Health Smart Home, Multimodal Corpora: Advances in Capturing, Coding and Analyzing Multimodality, LREC, 18-21 May 2010, Malta, pp. 99-105.

Abstract - Health Smart Homes are nowadays a very explored research area due to the needs for home automation and telemedicine to support people with loss of autonomy and also due to the evolution of the technology that has resulted in cheap and efficient sensors. However, %, collecting data in this area is still very challenging. As a consequence, many studies do not include tests in real setting, because data collection is this domain is very expensive and challenging and because of the few available data sets which are anyway difficult to reuse. In this article, we present a dataset acquired in real conditions during an experiment involving 15 participants who were performing several instances of seven activities in a fully equipped Health Smart Home.

Référence bibliographique : F. Portet, A. Fleury, M. Vacher, N. Noury, Determining Useful Sensors for Automatic Recognition of Activities of Daily Living in Health smart home, Intelligent Data Analysis in Biomedicine and Pharmacology, Verona, Italy, Jul. 19, 2009 Aug 20-24, 2009, pp. 63-64.

Abstract - To face the rapid growth of the world elderly population, health smart homes with sensing technology are emerging to automatically detect early loss of autonomy using objective criterion such as the Activity of Daily Living grid. The paper presents data mining techniques to classify 7 seven activities in a health smart home using only the most relevant attributes. The evaluation has shown that a correct classification of 84.5% can be reached with a dataset reduced to 16% related to less than 34% of the current sensors. Results also showed the importance of microphones as complementary data source.

Référence bibliographique : A. Fleury, N. Noury, M. Vacher, Application des SVM à la classification automatique des Activités de la Vie Quotidienne d'une personne à partir des capteurs d'un Habitat Intelligent pour la Santé, XVIèmes Rencontres de la Société Francophone de Classification, Grenoble, France, Sep. 2-4, 2009, 4p.

Abstract -

Référence bibliographique : A. Fleury, N. Noury, M. Vacher, Supervised Classification of Activities of Daily Living in Health Smart Homes using SVM, 31th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC09), Minnesota, USA, 2009, pp. 6099-6102.

Abstract - By 2050, about a third of the French population will be over 65. To face this modification of the population, the current studies of our laboratory focus on the monitoring of elderly people at home. This aims at detecting, as early as possible, a loss of autonomy by objectivizing criterions such as the international ADL or the French AGGIR scales implementing automatical classification of the different Activities of Daily Living. A Health Smart Home is used to achieve this goal. This flat includes different sensors. These data from the various sensors were used to classify each temporal frame into one of the activities of daily living that has been previously learnt (seven activities: hygiene, toilets, eating, resting, sleeping, communication and dressing/undressing). This is done using Support Vector Machines. We performed an experimentation with 13 young and healthy subjects to learn the model of activities and then we tested the classification algorithm (crossvalidation) on real data.

Key words: Support vector machines, Activities of Daily Living, Health Smart Homes, Sensor Fusion

Référence bibliographique : M., Vacher, A. Fleury, F. Portet, J.-F. Serignat and N. Noury, Reconnaissance des sons et de la parole dans un Habitat Intelligent pour la Santé : expérimentations en situation non contrôlée, GRETSI Traitement du signal et des images, Dijon, France, Sep. 8-11, 2009, 4p.

Résumé - L'article présente AUDITHIS, un système complet de reconnaissance des sons et de la parole, et son évaluation en environnement réel non contrôlé dans un appartement. Ce système permet de déclencher des alarmes suite à une détection de mots clés de détresse et apporte des informations audio utilisables pour la reconnaissance des Activités de la Vie Quotidienne (AVQ) d'une personne à domicile. Les expérimentations avec des participants jouant un scénario ont montré un taux de bonnes classifications de 8 classes de sons de 72% ainsi qu'un taux de fausse alarmes de 4% et un taux d'alarmes manquées de 30% pour la détection de mots clés de détresse. Concernant la reconnaissance des activités, des tests sur un ensemble de plus de 11 heures d'enregistrement sur 18 modalités ont été réalisés montrant l'intérêt des informations audio et apportant des perspectives d'amélioration.

Référence bibliographique : A. Fleury, N. Noury, M. Vacher, GRETSI Traitement du signal et des images, Traitement des signaux cinématiques pour la détection et la classification des transferts posturaux : le système ACTIM6D, Dijon, France, Sep. 8-11, 2009, 4p.

Résumé - Le faible prix des capteurs cinématique ainsi que leur faible consommation et facilité d'intégration ont permis de développer ces dernières années leur utilisation dans le cadre d'application biomédicales et notamment pour la surveillance des activités de la personne porteuse. Les travaux de recherche portent principalement sur la détection de certains moments de la vie quotidienne par reconnaissance de formes. Cet article décrit une méthode de détection des transferts posturaux utilisant la transformée en ondelettes afin de limiter le nombre de paramètres et de seuils à fixer pour des utilisations avec différentes populations. Cette méthode a ensuite été validée sur un scénario court donnant un taux de bonne classification global de 70%.

Référence bibliographique : M. Vacher, A. Fleury, N. Guirand, J.-F. Serignat,N. Noury, SPED 2009, Speech recognition in a smart home: some experiments for telemonitoring, From Speech Processing to Spoken Language Technology, Constanta, Romania, Jun. 18-21, 2009, pp. 171-179.

Abstract - Because of the aging of the population, low-cost solutions are required to help people with loss of autonomy staying at home rather than in public health centers. One solution is to assist human operators with smart information systems. In this case, position and physiologic sensors already give important information, but there are few studies about the utility of sound in patient's habitation. However, sound classification and speech recognition may greatly increase the versatility of such a system: this will be provided by detecting short sentences or words that could characterize a distress situation for the patient. Moreover, analysis and classification of sounds emitted in patient's habitation may be useful for patient's activity monitoring. In this paper, we present a global speech and sound recognition system that can be set-up in a flat. Eight microphones were placed in the health smart home of Grenoble (named HIS, a real living flat of 47m2) to automatically analyze and classify different sounds and speech utterances (e.g.: normal or distress French sentences). Sounds are clustered in eight classes but this aspect is not discussed in this paper. For speech signals, an input utterance is recognized and a subsequent process classifies it in normal or distress, by analysing the presence of distress keywords. An experimental protocol was defined and then this system has been evaluated in uncontrolled conditions in which heterogeneous speakers were asked to utter predetermined sentences in the HIS. The results of this experiment, where ten subjects were involved, are presented. The global error rate was 15.6%. Moreover, noise suppression techniques were incorporated in the speech and sound recognition system in order to suppress the noise emitted by known sources like TV or radio. An experimental protocol was defined and tested by four speakers in real conditions inside a room. Finally, we discuss the results of this experiment as a function of the noise source: speech or music.

Key words: Noise Suppression, Smart Home, Speech Recognition, Telemonitoring

Référence bibliographique : A. Fleury, N. Noury, M. Vacher, A wavelet-based pattern recognition algorithm to classify postural transitions in humans, 17th European Signal Processing Conference (EUSIPCO 2009), Glasgow, Scotland, 2009, 5p.

Abstract - Elderly people can be monitored at home to detect autonomy issues in their behavior. In addition to the environmental sensors (presence and movements in a room, temperature in the flat, light, etc.), we developed an inertial and magnetic based central of sensors to monitor the activity of the person. This article presents a wavelet-based pattern recognition algorithm that work on the data of this central to detect the postural transitions occurring in the daily life. We constructed four patterns: (1) for stand to sit, (2) for sit to stand, (3) for stand to lying down and (4) for lying down to stand. With this, we are able to detect theses changes, and to infer (as we consider that the subject is stand-up when the sensor is turned on) its posture. We also have, with these sensors, an idea of the activity of the person in each frame of time (immobile, moving, etc.). To test this algorithm and verify that the patterns are independent of the subject, we asked fifteen people to reproduce a scenario and we present, in the last section of this article, the results obtained. Results of an experiment are also given to show a mean good classification rate of 70% for this method.

Key words: Pattern recognition, Postural Transitions, Wavelet

Référence bibliographique : A. Fleury, M. Vacher, H. Glasson, J.-F. Serignat and N. Noury, Data Fusion in Health Smart Home: Primary Individual Evaluation of Two Families of Sensors, ISG 2008, the 6th International Conference of the International Society for Gerontechnology, Pisa, Italy, June 4-7 , 2008, pp. 6p.

Abstract -

Life expectancy is nowadays increasing thanks to major improvements in medicine. Thus, modern societies are facing the great challenge to care after a fast growing population of elderly people. For that reason, researchers work on solutions to maintain, as long as possible, elderly persons safely in their own home, with efficient systems to detect abnormal trends and to launch alarms. This project deals with the development of indicators to detect the loss of autonomy. A flat was equipped with different sensors to classify the activities of daily living of the patient in his own environment. This paper describes the installation of the different sensors and the results of the preliminary individual evaluation of two of them (a sound and speech recognition system and an inertial/magnetic kinematic sensor). The first system classifies the sounds produced in the flat in eight classes and differentiates the normal sentences with the distress sentences uttered by the subject. The second analyzes the signal from the accelerometers and magnetometers to identify the posture and the level of activity. The algorithms were tested with two scenarios performed by ten subjects.

Référence bibliographique : Michel Vacher, Anthony Fleury, Jean-François Serignat, Norbert Noury and Hubert Glasson, Preliminary evaluation of speech/sound recognition for telemedicine application in a real environment, INTERSPEECH, Brisbane, Australia, Sep 22-26, 2008, pp. 496-499.

Abstract -

Improvements in medicine increase life expectancy and the number of elderly persons, but the institutions able to welcome them are not sufficient. A lot of projects work on ways allowing elderly persons to stay at home. This article describes the implementation of a sound classification and speech recognition system equipping a real flat. This system has been evaluated in uncontrolled conditions for distinguishing normal sentences from distress ones; these sentences are uttered by heterogeneous speakers. The detected signals are first classified as sound and speech. The sounds are clustered in eight classes (object fall, doors clap, phone ringing, steps, dishes, doors lock, screams and glass breaking). As for speech signals, an input utterance (in French) is recognized and a subsequent process classifies it in normal or distress, by analysing the presence of distress key words. In the same way, some sound classes are related to a possible distress situation. An experimental protocol was defined and tested in real conditions inside the flat. Finally, we discuss the results of this experiment, where ten subjects were involved.

Référence bibliographique : A. Fleury, N. Noury, M. Vacher, H. Glasson and J.-F. Serignat, Sound and Speech Detection and Classification in a Health Smart Home, 30^th IEEE EMBS Annual International Conference, "Personalized Healthcare through Technology", Vancouver, British Columbia, Canada, Aug 20-24, 2008, pp. 4644-4647.

Abstract -

Improvements in medicine increase life expectancy in the world and create a new bottleneck at the entrance of specialized and equipped institutions. To allow elderly people to stay at home, researchers work on ways to monitor them in their own environment, with non-invasive sensors. To meet this goal, smart homes, equipped with lots of sensors, deliver information on the activities of the person and can help detect distress situations. In this paper, we present a global speech and sound recognition system that can be set-up in a flat. We placed eight microphones in the Health Smart Home of Grenoble (a real living flat of 47m2) and we automatically analyze and sort out the different sounds recorded in the flat and the speech uttered (to detect normal or distress french sentences). We introduce the methods for the sound and speech recognition, the post-processing of the data and finally the experimental results obtained in real conditions in the flat.

Référence bibliographique : M. Vacher, J.F.Serignat, S. Chaillol, Sound Classification in a Smart Room Environment: an Approach using GMM and HMM Methods, The 4^th Conference on Speech Technology and Human-Computer Dialogue, Advances in Spoken Language Technology, Iasi, Romania, 10-12 May, 2007, pp. 135-146, ISBN: 978-973-27-1516-1.

Abstract - Because of cost or convenience reasons, patients or elderly people would be hospitalized at home and smart information systems would be needed in order to assist human operators. In this case, position and physiologic sensors give already numerous informations, but there are few studies for sound use in patient's habitation. However, sound classification and speech recognition may greatly increase the versatility of such a system: this will be provided by detecting short sentences or words which could characterize a distress situation for the patient. Analysis and classification of sounds emitted in patient's habitation may be useful for patient's activity monitoring. GMMs and HMMs are well suited for sound classification. Until now, GMMs are frequently used for sound classification in smart rooms because of their low computational costs, but HMMs should allow a finer analysis: indeed the use of 3 states HMMs should allow better performances by taking into account the variation of the signal according to time. For this framework a new sound corpus was recorded in experimental conditions. This corpus includes 8 sound classes useful for our application. The choice of needed acoustical features and the two approaches are presented. Then an evaluation is made with the initial corpus and with additional experimental noise. The obtained results are compared. At the end of this framework a segmentation module is presented. This module has the ability of extracting isolated sounds in a record by the means of a wavelet filtering method which allows the extraction in noisy conditions.

Key words: Gaussian mixture model; Hidden Markov model; Background noise; Sound classification in smart rooms; Wavelet transform.

Référence bibliographique : D. Istrate, M. Vacher, J.F.Serignat, Embedded Implementation of Distress Situation Identification through Sound Analysis, the 5^th ICICTH Samos: International Conference on Information Communication Technologies in Health, 5-7 July 2007, Samos, Greece, pp. 226-231.

Abstract - The safety of elderly people living alone at home might be a crucial problem because of the growing aging population and the high risk of home accidents such as falls. Medical remote monitoring systems may increase the safety of such people in detecting and quickly announcing the state of emergency. We have already proposed a sound medical remote monitoring system. Distress sounds like glass breaking, screams, falls and distress expressions like "Help", "A doctor, please!" are detected and recognized through a continuous analysis of the sound flow. In the case of distress situation identification, the software can send an alarm with the recognized data to a close person and/or to a medical center. In this paper, a real-time implementation of this system is presented. The advantages of this implementation on an Embedded PC, equipped with a classical sound card and a microphone, are the reduced dimensions, the silence (fan less) and the cost. In the same time, this implementation is flexible and can be installed also on desktop or laptop PC.

Key words: sound detection, sound/speech classification, sound recognition, signal processing, embedded system, telemedicine.

Référence bibliographique : M. Vacher, D. Istrate, J.F.Serignat, Speech and Sound Analysis: an Application of Probabilistic Models, the International Symposium on System Theory, Automation, Robotics, Computers, Informatics, Electronics and Instrumentation (SINTES), 15-18 Octobre 2007, Craiova, Romania, pp. 173-178.

Abstract - Sound signal carry rich information which can be extracted and used by analysis systems in different modes. The medical remote monitoring systems may help elderly people to live home in security. We have already proposed an environmental sound analysis algorithm to detect distress sound or sentences. The sound and speech recognition modules based on Gaussian Mixture Models and Hidden Markov Models are detailed. A real time implementation on an embedded PC is proposed and evaluated. This implementation is flexible either for the hardware target (desktop, laptop or embedded PC) or on the alarm sending (E-Mail, SMS or remote monitoring center).

Key words: Signal processing algorithms, Gaussian distributions, Markov models, Pattern recognition, Real time computer systems, Embedded systems, Speech Analysis.

Référence bibliographique : M. Vacher, P. Menendez-Garcia, J.F.Serignat, D. Istrate, First Implementation of a Sound/Speech Remote Monitoring Real-Time System for Home Healthcare, presented at the 6^th International Conference IEEE Communications 2006, Bucharest (Romania), June 2006, pp.111-115.

Abstract - Medical Remote Monitoring needs human operator assistance by smart information systems. Physiological and position sensors give already numerous informations, but sound classification can give interesting additional informations about the patient and may help to the decision-making. A Real-Time implementation of a multichannel smart sound/speech system is presented in this paper, it is capable in a first step to detect and identify sound events in noisy conditions, in a second step to classify it into speech or life sound. According to that result the third step is speech recognition or sound classification. The multichannel sound processing allows us to localize the sound in the apartment and to select appropriate signal segments for identification procedure. Recognized sentences and classification results are sent to the medical remote monitoring application through Ethernet network. The system is composed of several parallel tasks: detection and channel selection, sound/speech classification, life sound classification, speech recognition and graphical interface. The event detection module is carried out for each channel in real time. The classification modules are launched in a parallel task. Speech Recognition is running as an independent application on the same computer.

The 6^th International Conference IEEE Communications 2006, Bucharest (Romania), 8-10 June 2006. Site de la conférence

Référence bibliographique : D. Istrate, M. Vacher, J.F.Serignat, Generic Implementation of a Distress Sound Extraction System for Elder Care, 28^th IEEE EMBS Annual International Conference, New York City, USA, Aug 30-Sept 3, 2006, pp.3309-3312.

Abstract - Medical remote monitoring at home is an alternative to improve the patient's comfort, to detect distress situation rapidly and reduce hospitalization costs. Physiologic and position sensors give already numerous information, but sound classification can give interesting additional information. A Real-Time implementation of a smart sound system capable of detecting and identifying sound events in noisy conditions is presented in this paper. The advantage of this implementation is the use of a generic PC station: the hardware requirements are only a sound card, a microphone and an internet link used to transmit alarm. In the case of an alarm, the information can be sent through network to a remote monitoring center and/or to a close person by email or SMS. The system is composed of 2 modules: detection and classification. The event detection module is carried out in real time in order to extract possible alarm sounds. The sound classification module is runs in a parallel task; it carries out a first segmentation between sound and speech. In the speech case, a speech recognition system is launched (not described in this paper) and in the sound case, a classification between predefined classes is carried out.

Référence bibliographique : M. Vacher, D. Istrate, J.F. Serignat and N. Gac, "Detection and SPeech/Sound Segmentation in a Smart Room Environment", presented at The 3rd Conference on Speech Technology and Human-Computer Dialogue (Sped 2005), Cluj-Napoca, Romania, May 13-14, 2005, pp. 37-48.

Abstract - Because of cost or convenience reasons, patients or elderly people would be hospitalized at home and smart information system would be needed in order to assist human operators. In this case, physiologic and position sensors give already numerous informations, but there are few studies for sound use in patient's habitation. However, sound classification and speech recognition may greatly increase the versatility of such a system: this will be provided by detecting specific sounds or short sentences which could characterize a distress situation for the patient. Sounds emitted in patient's habitation may be useful for patient's activity monitoring. The proposed sound analysis system is made of four modules: the first module in charge of sound and speech extraction is the detection module, it is followed by a segmentation module needed to transmit the extracted wave to the Sound Classification module or to the Speech Recognition module. The first two modules -Detection and Segmentation- are presented and evaluated in this paper in experimental recorded noise conditions. The detection method uses transient models, based upon dyadic trees of wavelet coefficients to insure short detection delay. The segmentation step is a classical Gaussian Mixture Model classifier based on acoustical parameters like MFCC.

Trends in Speech Technology, SPED 2005
The 3rd International Conference on Speech Technology and Human-Computer Dialogue, Cluj (Roumanie), 13-14 Mai 2005.

Référence bibliographique : D. Istrate, M. Vacher, J.F.Serignat, "Détection et classification des sons : application aux sons de la vie courante et à la parole", presented at GRETSI Traitement du signal et des image, Louvain-la-Neuve, Belgique, 6-9 septembre, 2005, vol. 1, pp. 485-488.

Résumé - Depuis quelques années se développe le concept général d'espace perceptif (salle intelligente) qui répond de diverses façons aux besoins, demandes, attentes des acteurs humains. Un système d'extraction de l'information du son à trois étapes est proposé. La première étape, permet la détection et l'extraction des sons du flux sonore continu. L'algorithme de détection proposé est basé sur la transformée en ondelettes, il permet de s'affranchir du bruit et d'obtenir une bonne résolution temporelle. La deuxième étape utilise un mélange de distributions de Gauss (GMM) pour faire la classification du signal sonore entre parole et sons et aiguiller le signal sur le processus adapté : reconnaissance de la parole (non traitée dans l'article) ou classification des sons. La troisième étape, celle de classification des sons de la vie courante, est aussi réalisée avec un système à base de GMM. Les paramètres acoustiques sont étudiés étant donné qu'ils ont une influence essentielle sur le système de classification ; par ailleurs, de nouveaux paramètres issus de la transformée en ondelettes sont proposés. Chaque étape de l'étude est validée au moyen d'un corpus spécifique.

Abstract - Recently, the general concept of perceptive spaces or smart rooms is in a continuous development and tries to answer in different ways to the needs, demands or expectations of human actors. This paper presents a system to extract information from sound signals, which contains three stages. The first stage, sound event detection, takes care of the sound detection and extraction from a continuous acoustic flux and it uses an algorithm based on the wavelet transform. The algorithm described in the paper offers good temporal resolution and performances in a noisy environment. The second stage aims to Speech/Sound classification and uses Gaussian Mixtures Models (GMM). The third stage is also based on GMM and realizes the classification of the everyday life sounds. The acoustical parameters are studied since they have an important influence on the classification performances. New parameters based on the wavelet transform are proposed. Every stage of the system is validated on a specific corpus.

Référence bibliographique : M. Vacher, D. Istrate, L. Besacier, J.F.Serignat and E. Castelli, "Sound Detection and Classification for Medical Telesurvey", presented at 2nd Conference on BIOMEDICAL ENGINEERING (BIOMED 2004), Innsbruck, Austria, February 16-18, 2004, pp. 395-399.

Abstract. - Medical Telesurvey needs human operator assistance by smart information systems. This paper deals with the sound event detection in a noisy environment and presents a first classification approach. Detection is the first step of our sound analysis system and is necessary to extract the significant sounds before initiating the classification step. An algorithm based on the Wavelet Transform is evaluated in noisy environment. Then Wavelet based cepstral coefficients are proposed and their results are compared with more classical parameters. Detection algorithm and sound classification methods are applied to medical telemonitoring. In our opinion, microphones surveying life sounds are better preserving patient privacy than video cameras.

2nd IASTED International Conference on BIOMEDICAL ENGINEERING
~BioMED 2004~ Innsbruck (Austria) 16-18 février 2004 (BioMED 2004) (International Association of Science and Technology for Development)

Référence bibliographique : M. Vacher, D. Istrate and J.F. Serignat, "Sound Detection and Classification through Transient Models using Wavelet Coefficient Trees", presented at 12th European Signal Processing Conference (EUSIPCO 2004), Vienna, Austria, September 6-10, pp. 1171-1174.

Abstract. - Medical Telesurvey needs human operator assistance by smart information systems. Usual sound classification may be applied to medical monitoring by use of microphones in patient's habitation. Detection is the first step of our sound analysis system and is necessary to extract the significant sounds before initiating the classification step. This paper proposes a detection method using transient models, based upon dyadic trees of wavelet coefficients to insure short detection delay. The classification stage uses a Gaussian Mixture Model classifier with classical acoustical parameters like MFCC. Detection and classification stages are evaluated in experimental recorded noise condition which is non-stationary and more aggressive than simulated white noise and fits with our application. Wavelet filtering methods are proposed to enhance performances in low signal to noise ratios.

12th European Signal Processing Conference
~EUSIPCO 2004~ Vienna (Austria) 6-10 septembre 2004 (EUSIPCO 2004)
(EURASIP)

Référence bibliographique : M. Vacher, D. Istrate and J.F. Serignat, "Sound Detection through Transient Models using Wavelet Coefficient Trees", presented at Complex Systems, Intelligence and Modern Technology Applications (CSIMTA 2004), Cherbourg, France, September 19-22, 2004, pp. 367-372.

Abstract. - Medical Telesurveillance needs human operator to be assisted by smart information systems. Therefore automatic determination of sound type emitted in patient's habitation may greatly increase the versatility of such a system. Sounds are acquired through microphones set out in each room. Detection is the first step of our sound analysis system and is necessary to extract the significant sounds before initiating the classification step. This paper proposes a detection method using transient models, based upon dyadic trees of wavelet coefficients to insure short detection delay. This method is used to detect at once the beginning and the end of the audio signal allowing signal extraction in noisy environment. The precision of this step is important to avoid a decrease of performances during the second step which is the classification step. This step uses a Gaussian Mixture Model classifier with classical acoustical parameters like MFCC. Detection and classification stages are evaluated in experimental recorded noise condition which is non-stationary and more realistic than simulated white noise. Wavelet filtering methods are proposed to enhance classification performances in low signal to noise ratios.

Complex Systems, Intelligence and Modern Technology Applications ~CSIMTA 2004~ Cherbourg (France) 19-22 septembre 2004 (CSIMTA 2004)

Référence bibliographique : D. Istrate, M. Vacher, J.F. Serignat and E. Castelli, "Multichannel Smart Sound Sensor for Perceptive Spaces", presented at Complex Systems, Intelligence and Modern Technology Applications (CSIMTA 2004), Cherbourg, France, September 19-22, 2004, pp. 691-696.

Abstract. - Sound based perceptive spaces are usually encountered for friendly man-machine interfaces, but sound information extraction for perceptive spaces is a complex task because of environmental noise and of multichannel processing need. A multichannel smart sound sensor capable to detect and identify sound events in noisy conditions is presented in this paper. The multichannel sound processing allows us to localize the sound in the perceptive space and to select appropriate signal for identification procedure. This sensor is real time implemented on PC. The event detection module is carried out for each channel in real time. The classification module is launched in a parallel task on the channel chosen by data fusion process. The aim of this process is to select the channel with the biggest signal to noise ratio when a multiple detection occurs. The validation of smart sensor is made on a test set and is presented with the proposed methodology of evaluation for a medical telemonitoring application. The obtained results are allowing us to develop perceptive space applications.

Complex Systems, Intelligence and Modern Technology Applications ~CSIMTA 2004~ Cherbourg (France) 19-22 septembre 2004 (CSIMTA 2004)

Référence bibliographique : D. Istrate, M. Vacher, E. Castelli and C. P. Nguyen, "Sound Processing for Health Smart Home", presented at 2nd International Conference on Smart Homes and Health Telematic, Singapore, September 15-17, 2004.

Abstract - Sounds in smart home are usually encoutered for friendly man-machine interfaces, but sound information is a complex task because of environmental noise and of multichannel processing need. A multichannel sound processing system capable to detect and identify sound events in noisy conditions is presented in this paper. The multichannel sound processing allows us to localize the sound in smart home and to select appropriate signal for identification procedure. This sensor is real time implemented on PC. The event detection module is carried out for each channel in real time. The classification module is launched in a parallel task on the channel chosen by date fusion process. The aim of this process is to select the channel with the biggest signal to noise ration when a multiple detection occurs. The system validation is made on a test set and is presented with the methodology of evaluation for a medical telemonitoring application. The obtained results are allowing us to develop smart home applications.

2nd International Conference on Smart Homes and Health Telematic, Singapore, 15-17 Septembre 2004.