Indian Journal of Science and Technology

Abstract

Objective: This paper investigates heart rate variability using neural networks. Methods: A software is developed to detect the heart rate variation using ECG signals. For this purpose the signal is divided into sub samples that are overlapping to a certain extent. The signals are transformed into the frequency domain using FFT (Fast Frequency Transform) and to decompose the signal, wavelet transform is used. Further the features are extracted using the above transforms which are fed into the neural networks. Findings: Heart rate variation is calculated using neural networks where the features are used to train them. For training, two learning algorithms are used, LM (Levenberg Marquardt) and BR (Bayesian Regularization). LM is found to converge faster than BR but the latter has higher efficiency. Application: Variation in heart rate can be used for better detection of diseases.

Keywords: Bayesian Regularization, Heart Rate Variability, Levenberg Marquardt Machine Learning, Signal Processing