Reproduction de : Thèse de doctorat : Acoustique : Le Mans : 2007. / Titre provenant de l'écran-titre. Bibliogr. p.203-209.

This work deals with possibilities of operating signals under the power supply voltage and using of this technology in guidance of a model of railway. Firstly there is an analysis of analogous and digital guidance of the model of railway. There is also a description of theory of three kinds of used modulations such as ASK, FSK and OFDM modulation. There are described only principles of those modulations because the problematic of the modulation is going to be solved by integrated circuit. The next part of this work is going to deal with a market research and with description integrated circuits, which ensure the modulation of operating signals of power supply voltage. There is also going to be noticed a description of others circuits that are needed for the correct function of the whole furnishing. Then I will continue with a project of scheme and single PCB for a guidance of the model of the railway. The last part of this work is concerned with activation and a proof of those proposed linkage.

Цель работы – исследование систем распознавания сигналов радиоэфира. / The aim of the work is to study the systems of recognition of radio signals.

This work presents a study of machine learning methods for classification of radio broadcast signals. The first chapter provides a brief overview of the subject area, as well as an analysis of current solutions on this topic in the field of machine learning, during which such models as LResNet and CCNN-Atten were considered. The second chapter described in detail the selected RadioML dataset and developed a structure of the experiments with the TimeSeriesForestClassifier, TimeForestClassifier, SVC, KNN models, taking into account various approaches to processing the original data. The third chapter of this work presents the results of the experiments: analysis of approaches to data processing, where the best solution was the calculation of the amplitude spectrum of the signal; results of selecting hyperparameters and testing models on an extended training set. The best identified model is TimeSeriesForestClassifier with an accuracy of 0.91 at an SNR of at least 10 dB. / В данной работе представлено исследование методов машинного обучения для классификации сигналов радиоэфира. В первой главе представлен краткий обзор предметной области, а также анализ актуальных решений на данную тематику в области машинного обучения, в ходе которого были рассмотрены такие модели как LResNet и CCNN-Atten. Во второй главе был подробно описан выбранный датасет RadioML и разработана структура проводимых экспериментов с моделями TimeSeriesForestClassifier, TimeForestClassifier, SVC, KNN с учетом различных подходов к обработке исходных данных. В третьей главе настоящей работы представлены результаты проводимых экспериментов: анализ подходов к обработке данных, где наилучшим решением оказалось вычисление амплитудного спектра сигнала; результаты подбора гиперпараметров и тестирования моделей на расширенной обучающей выборке. Наилучшей выявленной моделью является TimeSeriesForestClassifier с accuracy 0.91 при SNR не менее 10 dB.

This thesis deals with generalizations of the analytic signal (AS) construction proposed by Gabor. Functional extensions of the fractional Hilbert Transform (FrHT) are proposed using which families of analytic signals are obtained. The construction is further applied in the design of a secure communication scheme. A demodulation scheme is developed based on the generalized AS, motivated by perceptual experiments in binaural hearing. Demodulation is achieved using a signal and its arbitrary phase-shifted version which, in turn translated to demodulation using a pair of flat-top bandpass filters that form an FrHT parir. A new family of wavelets based on the popular Gammatone auditory model is proposed and is shown to lead to a good characterization of singularities/transients in a signal. Allied problems of computing smooth amplitude, phase, and frequency modulations from the AS. Construction of FrHT pair of wavelets, and temporal envelope fit of transient audio signals are also addressed.