Prediction of Exhaustion Threshold Based on ECG Features Using The Artificial Neural Network Model

Awareness on physical fatigue is important for prevention of injury caused by over-exercise while performing an exercise. Therefore, this study aims to predict an exhaustion threshold quantitatively based on ECG features using the artificial neural network. All five males of volunteered were participated to obtain the exhaustion threshold from exercise-induced fatigue protocol. By using the electrocardiogram (ECG) device, the electrodes were attached on the thorax to comply with the Einthoven's Law. Also, it equipped with Bluetooth for the data transmission for the real-time monitoring. Then, the ECG raw signal was filtered to eliminate the noise by using the 4th order Butterworth filter in low-pass frequency. Statistic feature extraction in each window provides specific information for the input data while the Borg scale and time to exhaustion are target data. Subject 2 and 3 contributed the highest time of exhaustion (TTE) with minimum HR reserve and categorised as normal BMI. Only 56 input data were considered as optimum parameters for the fastest time and regression value t=0.7313s and R=0.99668 respectively. The regression value for both outputs Borg's scale and TTE present the significant correlations by produced R=0.99095 and R=0.99541. The highest threshold prediction contributed up to 89.3%, to the good mathematical model developed. This method is promising for the prediction of exhaustion threshold in order to replace the qualitative with the quantitative measurement. Therefore, it is beneficial for the athlete as well as coach to monitor exhaustion and manage it properly from severe injury happen.