Investigation Reinforcement Learning Method for R-Wave Detection on Electrocardiogram Signal

One of the essential things in recording electrocardiography (ECG) signals is R-wave detection. R-wave is a main component in the QRS wave, where it used for the diagnosis of heart rhythm irregularities and to calculate heart rate variability. On the other hand, Reinforcement Learning (RL) method has been proved that gives an excellent performance for games, network and stock market prediction. Based on this fact, in this paper, we want to investigate the RL for R-wave detection on the ECG signal. For the detection, the lower frequency is removed then we calculate the candidate of the peak, initialize Q-State-Action (QSA) table, peak detection, calculate reward and update the QSA table, respectively. Based on the experiment using the MIT-BIH dataset, the best accuracy of the proposed method is 86.8%, which calculated from alpha and gamma parameters are 0.1 and 0.9, respectively.