Розпізнавання медико-біологічних сигналів за ізолініями в системах, що навчаються з учителем
| dc.contributor.advisor | Шуляк, Олександр Петрович | |
| dc.contributor.author | Безнос, Денис Вікторович | |
| dc.date.accessioned | 2023-08-29T13:08:48Z | |
| dc.date.available | 2023-08-29T13:08:48Z | |
| dc.date.issued | 2021 | |
| dc.description.abstract | Мета роботи – розробка програмного інструментарію формування додаткової системи ознак опису форми медико-біологічних сигналів різних типів як ознак для їх розпізнавання. Об'єктом досліджень і розробок є ізолінії сигналів як додаткові системи ознак для їх розпізнавання. Предметом розробок є програмні процедури формування та використання систем ізоліній медико-біологічних сигналів для їх розпізнавання в системах, які навчаються з учителем. Зміст розробок розкривається на прикладі розпізнавання типів QRS-комплексів ЕКГ пацієнта. Актуальність теми роботи пов'язана з тим, що кожна система ознак відзначає свої особливості у формі сигналів. Ізолінії як додаткові ознаки дають більш повний облік форми сигналів і, можливо, підвищать якість їх розпізнавання. У розділі 1 роботи розглядається зміст завдання, яке в ньому вирішується, пропонуються принципи вирішення основних питань. Розділ 2 розкриває зміст і результати розробки необхідних програмних процедур. Розділ 3 присвячений питанням їх статистичного дослідження. Робота доповідалась на 25-му Міжнародному молодіжному форумі «Радіоелектроніка і молодь у ХХІ столітті» 2021 та XX Міжнародній науково- технічній конференції “Приладобудування: стан і перспективи”, 2021. Результати корисні у вдосконаленні процедур розпізнавання медико-біологічних сигналів. | uk |
| dc.description.abstractother | The topic of Diploma work is: «Recognition of biomedical signals by isolines in supervised learning systems.». In general, the work is aimed at improving software procedures for recognizing biomedical signals in patient diagnostic systems. The purpose of the work is to develop software tools for the formation of an additional system of signs describing the form of biomedical signals of different types as signs for their recognition. In addition to the sequence of samples shape characteristic of signals, systems of isolines to these characteristics are formed on equable time grid, which provide an additional numerical description of these characteristics. In this work, we consider the question of using only such additional descriptions of the shape characteristics of signals in recognition supervised learning systems The object of research and development is signal isolines as additional feature systems for their recognition. The subject of the development is software procedures for the formation and use of systems of isolines of biomedical signals for their recognition in supervised learning systems The substance and results of the developments are revealed on an illustrative example of recognizing the types of QRS-complexes in the recording of the patient's electrocardiogram. The relevance of the topic of the work is related to the fact that each system of features that describe the signal form notes its own features in the signal form, and the formation of additional features provides a more complete account of the features of the signal form in their recognition and this can become the basis for revealing possible reserves in improving its effectiveness. In the first section of the work, the content of the task of developing program procedures for recognizing biomedical signals in the supervised learning systems is determined, an approach to its solution is developed, the procedure for its implementation and the procedure for researching the features of using such procedures are determined. Solution of these questions is revealed on the illustrative example of determining the types of QRS complexes of the patient's electrocardiogram, which has the necessary marking by the position and types of these complexes in the record and a significant number of them for training the recognition algorithm. In order to solve the problem of developing procedures for recognizing biomedical signals using their isolines, the principles of solving such particular problems were defined: - the principle of sequential stepwise formation of systems of isolines of a predetermined order with its increase to the value set by the researcher. The isolines of the previous level are combined with new isolines of the current order for each instance of the signal whose shape is characterized. The isoline levels describe the shape characteristic of each such instance; - the principle of the coordinated connection of the mentioned systems of isolines for the formation of the resulting systems of levels for the description form of signals, which are used in the work as features for determining the types of these signals; - the principle of iterating over the intervals on the scale shape characteristics of signals to determine one isoline in each of them; - the principle of determining the borders of intervals in their sequence in the process of iterating from large values of the shape characteristic to smaller; - the principle of selection of samples of the shape characteristic of signal, which are taken into account in the calculation of the isoline level in each such interval of the scale of values shape characteristic of signals; - the principle and formulas of samples shape characteristics of signals in the calculation of its new isolines in each interval of the scale shape characteristics of signals; To implement each of these principles of solving the corresponding particular problems in the construction of systems of isolines of an instance of a signal, block diagrams of software modules are proposed, the development of which is discussed in section 2 in MATLAB. Such flowcharts are combined into the final flowchart of the procedure for obtaining the system of isolines of the signal instance. The necessary database has also been prepared for obtaining training and control samples of signals by their types. Section 2 of the work presents the development of software procedures for recognizing biomedical signals using their isolines in supervised learning systems. The development of such basic procedures is presented: 1. The procedure for converting the signal to a characteristic of its shape, which is carried out at the prior stage of processing the patient's ECG recording. A sequence of samples on a uniform time grid in units of measurement of the signal is converted on the same time grid into a sequence of dimensionless samples of the characteristic of its shape. 2. Final software procedure for forming a system of isolines of a signal instance, which are used as signs of its shape for recognition. The isolines are obtained as part of a system of levels of the required order, which is supplemented by service levels: the values of the upper and lower levels, and also the zero level of the scale shape characteristic of the signal. This procedure includes: - procedure for viewing intervals on the scale of signal instance shape characteristics and forming their borders; - module for selecting and counting samples shape characteristics of signal in the calculation of each regular isoline inside for their current interval; - a software module for combining new isolines with levels of the previous order. 3. Procedure for generating signal descriptions using isolines as part of a training and control sample. The sequence of isoline levels for the shape characteristic of the signal is converted into a set of numbers (vector) for the shape characteristic of this sequence. Such vectors are used as numerical feature systems for recognizing signals using a scalar product for comparing them in the decision-making process. 4. Procedure for forming etalons (characteristic descriptions of signal types) based on training samples using isolines. The procedure provides an average view of such descriptions for each type of signal. 5. Procedure for recognizing signals based on their isolines as features for recognition. The procedure calculates the inner products of the signal description vectors. The decision is made in favor of the signal type for which the inner product has the greatest value. The implementation of the program procedures was performed in the MATLAB for prior studies of the effectiveness of using the proposed features of the description of signals for their recognition The purpose of the final section of the work was to pre-test the readiness of the developed software procedures for their use and to determine the procedure for evaluating the quality of signal recognition by standard indicators in further research. Such particular problems were solved: 1. Defined: purpose, objectives, data, conditions and procedure for conducting numerical research on conducting statistical studies of signal recognition using their isolines as features for an illustrative example of signals. 2. Solved: content of the task and indicators for evaluating the quality of signal recognition using their isolines as features for recognition for an illustrative example of signals. As such indicators, standard indicators of the quality of the recognition algorithms are used. Such indicators in the work are the sensitivity, specificity, and overall validity of solutions, which show the recognition procedures in the process of checking them on control samples after they are trained on training samples. There are 62 instances of signals in each of their three types, both in the training and control samples. 3. The software procedure for forming the shape characteristic of the signal instance that is recognized is configured. 4. The software procedure for generating the isoline system for the instance of the characteristic of the recognized signal shape is configured. 5. The procedure for conducting numerical statistical studies of the quality indicators of signal recognition using isolines as their features is defined. In debugging the procedure for forming the characteristic of the signal form, the correspondence of the characteristic form and the signal, the absence of a constant component in the array of its values, and the unit length for the vector of samples of the characteristic were checked. The possibility of its use at the stage of prior transformation of the signals of the training and control samples in the process of obtaining additional descriptions of their signals using isolines is confirmed. Debugging of the software procedure for the formation of the system of isolines for the characteristic of the signal form was carried out by the stages of its formation in the process of processing individual instances of the sample signals. The correctness of dividing the samples shape characteristics of signals into subsets was checked in the process of increasing the order of the isoline systems and changing their composition, the correctness of the placement of the isolines among themselves in their general system was confirmed, and the presence of possible significant differences in the values of the levels that they hold on the scale of shape characteristic of signals for different types of signals was demonstrated. It is shown that the usefulness of such additional features in the recognition of signals by their shape is not excluded. In determining the procedure for conducting statistical researches of signal recognition quality indicators using isolines as their features, a flowchart of the corresponding procedure is constructed, and the necessary formulas for calculations are given. The sensitivity, specificity, and overall validity of signal recognition are provided as indicators of the quality of signal recognition. The content of the work was reported on XXV International Youth Forum «Radio electronics and youth in the XXIst century» 2021 and XX International Scientific and Technical Conference " INSTRUMENT ENGINEERING: state and prospects” 2021 The results of the work can be useful in research on improving software procedures for recognizing biomedical signals in patient diagnostic systems. | uk |
| dc.format.extent | 94 с. | uk |
| dc.identifier.citation | Безнос, Д. В. Розпізнавання медико-біологічних сигналів за ізолініями в системах, що навчаються з учителем : дипломна робота … бакалавра : 153 Мікро- та наносистемна техніка / Безнос Денис Вікторович. – Київ, 2021. – 94 с. | uk |
| dc.identifier.uri | https://ela.kpi.ua/handle/123456789/59624 | |
| dc.language.iso | uk | uk |
| dc.publisher | КПІ ім. Ігоря Сікорського | uk |
| dc.publisher.place | Київ | uk |
| dc.subject | діагностичні системи | uk |
| dc.subject | медико-біологічні сигнали | uk |
| dc.subject | розпізнавальні процедури | uk |
| dc.subject | опис форми сигналів ізолініями | uk |
| dc.title | Розпізнавання медико-біологічних сигналів за ізолініями в системах, що навчаються з учителем | uk |
| dc.type | Bachelor Thesis | uk |
Файли
Контейнер файлів
1 - 1 з 1
Вантажиться...
- Назва:
- Beznos_bakalavr.pdf
- Розмір:
- 2.13 MB
- Формат:
- Adobe Portable Document Format
- Опис:
Ліцензійна угода
1 - 1 з 1
Ескіз недоступний
- Назва:
- license.txt
- Розмір:
- 9.1 KB
- Формат:
- Item-specific license agreed upon to submission
- Опис: