Remoção da interferência da rede elétrica em sinais ecg baseado em análise de componentes independentes

• Autor(a) principal: Freire Júnior, Letivan Cambraia
• Data de Publicação: 2020
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFLA
• Texto Completo: http://repositorio.ufla.br/jspui/handle/1/56314
• topic: Eletrocardiograma; Análise de Componentes Independentes; Interferência da Rede Elétrica; Biomedical Signals; Independent Component Analysis; Power Line Interference; Automação Eletrônica de Processos Elétricos e Industriais

Biomedical signals are records of the activities of the human body. An example of a biomedical signal is the electrocardiogram (ECG), which is the recording of electrical oscillations caused by cardiac muscle activities. The ECG can help in the diagnosis and prevention of cardiac diseases, such as myocardial infarction and arrhythmia. These signals are subject to the appearance of several noises in your recording, due to the patient’s breathing and interference from the electrical network, for example. The latter is caused mainly due to the presence of electric and magnetic fields in the ECG recording environment, due to the use of electrical equipment, failures in the equipment grounding, among others. This leads to signal degradation, which makes it difficult to analyze, and it is necessary to employ techniques that attenuate these noises.In this work, a method for removing noise from the electrical network is proposed. The method is based on the use of digital FIR filters, Independent Component Analysis (ICA) and Fast Fourier Transform (FFT). Filters are used to generate sufficient observation signals for the application of the ICA. After its application, the independent sources contained in the ECG signal are extracted. The sources corresponding to the electrical network noise and its harmonic components are identified and are eliminated after the FFT processing. Finally, the sources estimated by the signal are used for signal reconstruction. To develop the method, ECG signals obtained from the PhysioBank ATM platform and the MIT-BIH Arrythimia databases were used, where the noise was added manually, and Challenge 2011 Training SET A, which has signals originally corrupted by various noises. The method showed good results and proved to be robust to variations that may occur, such as the presence of harmonic components of the electrical network signal and oscillations in the signal-to-noise ratio of the original signal with the added noise. The results were compared with a notch filter and an adaptive filter with LMS algorithm, showing better performance. All simulations and procedures were performed in the MatLab programming environment.