Novel time-frequency features for normal ECG

Literature was reviewed on the current state of methods for detecting arrhythmias. a software package,heartfelt,was developed that automatically detects arrhythmias from the lead ii ecg waveforms. to this end time-frequency distributions for simulated normal and abnormal ecg waveforms were obtained. then features representing the normal waveform were extracted and used to train a classifier. finally the performance of the classifier in detecting arrhythmias was tested using unknown ecg waveforms. the execution time of heartfelt was found to be 0.92 seconds for each minute of ecg recording. it was determined that the accuracy was 96.7% for detecting normal ecg signals,98.2% for detecting arrhythmic ecg signals suffering from arterial fibrillation,and 96.5% for detecting arrhythmic ecg signals suffering from superventricle arrhythmias.