Novel electrocardiogram segmentation algorithm using a multiple model adaptive estimator

Abstract

This thesis presents a novel electrocardiogram (ECG) processing algorithm design based on a Multiple Model Adaptive Estimator (MMAE) for a physiological monitoring system. Twenty ECG signals from the MIT ECG database were used to develop system models for the MMAE. The P-wave, QRS complex, and T-wave segments from the characteristic ECG waveform were used to develop hypothesis filter banks. By adding a threshold filter-switching algorithm to the conventional MMAE implementation, the device mimics the way a human analyzer searches the complex ECG signal for a useable temporal landmark and then branches out to find the other key wave components and their timing. The twenty signals and an additional signal from an animal exsanuinaiton experiment were then used to test the algorithm. Using a conditional hypothesis-testing algorithm, the MMAE correctly identified the ECG signal segments corresponding to the hypothesis models with a 96.8% accuracy-rate for the 11539 possible segments tested. The robust MMAE algorithm also detected any misalignments in the filter hypotheses and automatically restarted filters within the MMAE to synchronize the hypotheses with the incoming signal. Finally, the MMAE selects the optimal filter bank based on incoming ECG measurements. The algorithm also provides critical heart-related information such as heart rate, QT, and PR intervals from the ECG signal. This analyzer could be easily added as a software update to the standard physiological monitors universally used in emergency vehicles and treatment facilities and potentially saving thousands of lives and reducing the pain and suffering of the injured. Subject Terms Physiological monitoring, vital-sign monitoring, MMAE, ECG, Electrocardiogram Report Classification unclassified Classification of this page unclassified Classification of Abstract unclassified Limitation of Abstract UU Number of Pages

Algoritmos genéticos aplicados a la optimización de características en la clasificación de arritmias cardiacas utilizando los clasificadores KNN y naive Bayes

“…Hoffman et al [25] presentan un algoritmo de procesamiento de ECG basado en Multiple Model Adaptive Estimator (MMAE) para un sistema de vigilancia fisiológica, utilizando veinte señales de la base de datos de ECG del MIT.…”

Speech signal analysis for the estimation of heart rates under different emotional states

“…There had been different machine learning algorithms, namely, Support Vector Regression (SVR), Support Vector Machines (SVM), Sequential Minimal Optimization (SVR), and binary classification models, implemented. Additionally, there are many Low Level Descriptors (LLD) for voice signal feature extraction, and statistical functions, which are available in openSMILE software [25], [26]. However, they do not explicitly show, which feature extraction methods and algorithms are constituting to the estimation of factors between HR and human speech, [27].…”