|
| Titre : | Arial fibrillation episodes recognition using stockwel transform and support vector machines | | Type de document : | theses et memoires | | Auteurs : | Hocine Hamil ; Zahia Zidelmal, Directeur de thèse | | Editeur : | Tizi.Ouzou : U.M.M.T.O | | Année de publication : | 2021 | | Importance : | 192 p. | | Présentation : | ill. | | Format : | 30 cm | | Note générale : | Bibliogr. | | Langues : | Anglais | | Mots-clés : | ECG Atrial Fibrillation S-Transform Time-Frequency Analysis SVM e-Health Chaos Artificial Intelligence Diagnosis Wireless Transmission Biomedical Signal Processing Machine Learning | | Résumé : | The objective of this thesis is the detection of atrial fibrillation episodes using Stockwell transform and support vector machines. The first part concerns the implementation of the Stockwell transform with a compact support kernel (ST-CSK) for the localization of atrial activity. This zone will then be captured by a vector of features defined in
the time-frequency domain. The second part concerns the classification of the different atrial activities using cost-sensitive SVM where ambiguity rejection and cost of classification are considered allowing the recognition of subjects with atrial fibrillation or at risk of atrial fibrillation. The proposed algorithm is trained and then tested using different ECG sources namely two databases provided by PhysionNet and recorded ECG signals using MySignals HW development platform with raspberry Pi 3 model B+. The obtained results confirm that the proposed approach represents a promising tool for
Atrial Fibrillation Episodes (AFE) recognition even under real and clinical conditions.
Also in this context, a secure telehealth system with a wireless transmission has been designed analysing multiple bio-signals using e-Health platform with Arduino Uno and Raspberry Pi as acquisition and processing units, respectively. Data acquiered from different sensors can be evaluated to monitor the health status of patients using threshold detection. Prediction of cardiac status based on automatic identification of arrhythmias is validated by classification of ECG signals using Artificial Intelligence (AI), while exploiting TensorFlow and Keras tools. Different AI algorithms and a combination with different machine learning (ML) are validated basing to the transfer learning approach. The proposed method was evaluated using real signals recorded in addition to four PhysioNet databases. A graphical user interface (GUI) was designed allowing to display and to share the diagnosis with the prediction results. The proposed system improves the quality of life under a robust secure wireless transmission using chaos-based data encryption, thus avoiding loss of life during critical situations | | Diplôme : | Doc | | En ligne : | D:\CD THESES 2021\DOC ELN\HAMIL HOCINE.PDF | | Format de la ressource électronique : | PDF | | Permalink : | ./index.php?lvl=notice_display&id=36612 |
Arial fibrillation episodes recognition using stockwel transform and support vector machines [theses et memoires] / Hocine Hamil ; Zahia Zidelmal, Directeur de thèse . - Tizi.Ouzou (Tizi.Ouzou) : U.M.M.T.O, 2021 . - 192 p. : ill. ; 30 cm. Bibliogr. Langues : Anglais | Mots-clés : | ECG Atrial Fibrillation S-Transform Time-Frequency Analysis SVM e-Health Chaos Artificial Intelligence Diagnosis Wireless Transmission Biomedical Signal Processing Machine Learning | | Résumé : | The objective of this thesis is the detection of atrial fibrillation episodes using Stockwell transform and support vector machines. The first part concerns the implementation of the Stockwell transform with a compact support kernel (ST-CSK) for the localization of atrial activity. This zone will then be captured by a vector of features defined in
the time-frequency domain. The second part concerns the classification of the different atrial activities using cost-sensitive SVM where ambiguity rejection and cost of classification are considered allowing the recognition of subjects with atrial fibrillation or at risk of atrial fibrillation. The proposed algorithm is trained and then tested using different ECG sources namely two databases provided by PhysionNet and recorded ECG signals using MySignals HW development platform with raspberry Pi 3 model B+. The obtained results confirm that the proposed approach represents a promising tool for
Atrial Fibrillation Episodes (AFE) recognition even under real and clinical conditions.
Also in this context, a secure telehealth system with a wireless transmission has been designed analysing multiple bio-signals using e-Health platform with Arduino Uno and Raspberry Pi as acquisition and processing units, respectively. Data acquiered from different sensors can be evaluated to monitor the health status of patients using threshold detection. Prediction of cardiac status based on automatic identification of arrhythmias is validated by classification of ECG signals using Artificial Intelligence (AI), while exploiting TensorFlow and Keras tools. Different AI algorithms and a combination with different machine learning (ML) are validated basing to the transfer learning approach. The proposed method was evaluated using real signals recorded in addition to four PhysioNet databases. A graphical user interface (GUI) was designed allowing to display and to share the diagnosis with the prediction results. The proposed system improves the quality of life under a robust secure wireless transmission using chaos-based data encryption, thus avoiding loss of life during critical situations | | Diplôme : | Doc | | En ligne : | D:\CD THESES 2021\DOC ELN\HAMIL HOCINE.PDF | | Format de la ressource électronique : | PDF | | Permalink : | ./index.php?lvl=notice_display&id=36612 |
|
Arial fibrillation episodes recognition using stockwel transform and support vector machines (2021) 
