The detection of estrus continues to be a primary factor contributing to poor reproductive performance in modern dairy cattle. The objectives of this research were 1) to evaluate performance of automated detection of estrus using a reference standard of ovulation detection with temporal progesterone patterns 2) to evaluate the efficacy of parameters measured by automated detection of estrus systems 3) to evaluate the efficacy of alerts generated by several commercially available systems used for automated detection of estrus and 4) to determine the differences in these parameters among cows with or without poor health conditions at the time of estrus. Systems used for automated detection of estrus can perform better than the previous original reference standard, visual observation for standing behaviors. All systems used for automated detection of estrus tested were similar for estrus detection efficiency.

The main objective of this thesis is the development of new processing techniques for digital holograms. The present work is part of the HoloFlow project that intends to integrate the DHM technology for the monitoring of water quality. Different tools for an automated analysis of digital holograms have been developed to detect, refocus and classify particles in continuous fluid flows. A detailed study of the refocusing criterion permits to determine its dependencies and to quantify its robustness. An automated detection procedure has been developed to determine automatically the 3D positions of organisms flowing in the experiment volume. Two detection techniques are proposed: a usual method based on a global threshold and a new robust and generic method based on propagation matrices, allowing to considerably increase the amount of detected organisms (up to 95 %) and the reliability of the detection. To handle the case of aggregates of particles commonly encountered when working with large concentrations, a new separation procedure, based on a complete analysis of the evolution of the focus planes, has been proposed. This method allows the separation aggregates up to an overlapping area of around 80 %. These processing tools have been used to classify organisms where the use of the full interferometric information of species enables high classifier performances to be reached (higher than 93 %). / Doctorat en Sciences de l'ingénieur / info:eu-repo/semantics/nonPublished

Atrial fibrillation is an arrhythmia commonly detected from ECG using its specific characteristics. An early detection of this arrhythmia is a key to prevention of more serious conditions. Nowadays, atrial fibrillation detection is being implemented more often using deep learning. This work presents detection of atrial fibrillation from 12lead ECG using deep convolutional network. In the first section, there is a theoretical context of this work, then there is a description of proposed algorithm. Detection is implemented by a program in Python in two variations and their accuracy is rated by Accuracy and F1 measure. Results of the work are being discussed, mutually compared and compared to other similar publications.

Cílem této práce byl vývoj programu pro automatickou detekci arousalu v signálu spánkového EEG s použitím metod časově-frekvenční analýzy. Předmětem studie bylo 13 celonočních polysomnografických nahrávek (čtyři svody EEG, EMG, EKG a EOG), tj. celkově více než 100 hodin záznamu. Jednalo se o část dat z dřívějších výzkumných prací expertní lékařky v problematice spánku Dr. Emilie Sforzy, Ženeva, Švýcarsko, která rovněž poskytla základní hodnocení těchto dat. V záznamech bylo celkem označeno 1551 arousal událostí. Pro usnadnění výběru konkrétní metody časově-frekvenční analýzy byla následně vytvořena sada nástrojů pro vizualizaci jednotlivých signálů a jejich různých časově-frekvenčních vyjádření. S ohledem na závěry vizuální analýzy, charakter signálu EEG a efektivitu výpočetních metod byla pro analýzu vybrána waveletová transformace s mateřskou vlnkou Daubechies řádu 6. Jednotlivé svody EEG byly dekomponovány do šesti frekvenčních pásem. Z takto odvozených signálů a signálu EMG byly následně stanoveny ukazatele možné přítomnosti události arousalu. Tyto ukazatele byly dále váhovány lineárním klasifikátorem, jehož hodnoty vah byly optimalizovány pomocí genetického algoritmu. Na základě hodnoty lineárního klasifikátoru bylo rozhodnuto o přítomnosti události arousalu v daném svodě EEG – arousal byl detekován, jestliže hodnota klasifikátoru překročila danou mez na dobu více než 3 a méně než 30 vteřin. V celém záznamu pak byl arousal označen, byl-li detekován alespoň v jednom ze svodů EEG. Následně byly odvozeny míry senzitivity a selektivity detekce, jež byly rovněž základem pro stanovení fitness funkce genetického algoritmu. Pro učení genetického algoritmu byly vybrány první čtyři záznamy. Na základě takto optimalizovaných vah vznikl program pro automatickou detekci, který na celém souboru 13 záznamů dosáhl ve srovnání s expertním hodnocením míry senzitivity 76,09%, selektivity 53,26% a specificity 97,66%.

La baleine bleue Antarctique, Balaenoptera musculus intermedia, est en danger critique d’extinction depuis la chasse baleinière intensive du 20e siècle. L’état de ses populations et leur écologie restent encore mal connus. En raison de l’inefficacité des observations visuelles, la surveillance par acoustique passive est privilégiée pour étudier cette espèce vocalement très active. Cette thèse porte sur l’analyse de 7 ans de surveillance acoustique passive dans l’océan Indien austral, région d’habitat et de migration particulièrement importante pour la baleine bleue Antarctique. Déployé depuis 2010 sur une aire de près de 9 000 000 km2, le réseau d’hydrophones OHASISBIO fournit une base de données acoustiques multi-site et pluri-annuelle. L’application d’un algorithme de détection automatique des vocalisations de baleines bleues Antarctique, préalablement testé et validé, a permis d’établir les patrons géographiques et saisonniers de présence de l’espèce au sein du réseau. L’analyse systématique de ces vocalisations a également permis de caractériser des variations intra- et inter-annuelles de leur fréquence, affectée par une décroissance long-terme et des modulations saisonnières. L’analyse préliminaire de signatures vocales d’autres espèces présentes dans le réseau - rorquals communs et trois populations de baleines bleues pygmées – a révélé des variations de fréquence similaires de leur vocalisation et permis d’esquisser leurs patrons géographiques et saisonniers. Enfin, deux vocalisations, jusqu’alors non décrites, aux caractéristiques semblables à celles de baleines bleues, ont été identifiées et caractérisées. / The Antarctic blue whale, Balaenoptera musculus intermedia, is currently critically endangered since the commercial whaling in the 20th century. The population recovery of this species, as well as its ecology, are still poorly known. Due to the ineffectiveness of visual observations, passive acoustics is a preferred method to monitor this highly vocal species. This dissertation presents an analysis of 7 years of passive acoustic monitoring in the southern Indian Ocean, known as a particularly important area of habitat and migration for the Antarctic blue whale. Deployed since 2010 over an area of about 9,000,000 km2, the OHASISBIO hydrophone network provides a multi-site and multi-year acoustic database. An algorithm for the automated detection of Antarctic blue whale calls, first tested and validated, has been applied to characterize the seasonal and geographic patterns of the species presence in the study area. The systematic analysis of these vocalizations also allowed to characterize intra- and inter-annual variations of their frequency, with a long-term decline and seasonal variations. A preliminary analysis of other vocal signatures recorded by the network, from 3 populations of pygmy blue whales and fin whales, highlighted similar variations of their frequencies and outlined their geographic and seasonal patterns of presence in the area. Finally, two previously undescribed vocalizations, with characteristics close to that of blue whale calls, were identified and characterized.

The task of the judge of difficulty in trampoline gymnastics is to check the elements and difficulty values entered on the competition cards and the difficulty of each element according to a numeric system. To do this, the judge must count all somersaults and twists for each jump during a routine and thus record the difficulty of the routine. This assessment can be automated with the help of inertial measurement units (IMUs) and facilitate the judges’ task during the competition. Currently, there is no known reliable method for the automated detection and recognition of the various elements to determine the difficulty of an exercise in trampoline gymnastics. Accordingly, a total of 2076 jumps and 50 different jump types were recorded over the course of several training sessions. In the first instance, 10 different jump types were used to train different machine learning (ML) models. Eight ML models were used for the automatic jump classification. Supervised learning approaches include a naive classifier, deep feedforward neural network, convolutional neural network, k‑nearest neighbors, Gaussian naive Bayes, support-vector classification, gradient boosting classifier, and stochastic gradient descent. When all classifiers were compared for accuracy, i.e., how many jumps were correctly detected by the ML model, the deep feedforward neural network and the convolutional neural network provided the best matches with 96.4 and 96.1%, respectively. The findings of this study will help to develop the automated classification of sensor-based data to support the judge and, simultaneously, for automated training logging.