A Deep Learning Method to Translate Phonocardiogram (PCG) Signals to Electrocardiogram (ECG) Waveforms

In clinical practice, an electrocardiogram (ECG), which analyzes the electrical activity of the human heart, is a common cardiovascular diagnostic test. The ECG data can be stored and diagnosed effectively. However, precise positioning of the ECG electrodes is necessary for a reliable ECG tracing. This has an impact on how ECG monitors are used to check on patients at home. On the other hand, the Phonocardiogram (PCG) is a standard method for quantifying the properties of heart sounds in order to monitor and assess cardiac activity. With the advancement of wearable sensors, it is now simpler to record the heart's sound. We proposed a deep learning approach to translating PCG signals to ECG waveforms. This technique can be utilized to monitor cardiovascular disease in consumer wearable device development. The proposed method is evaluated on a benchmark dataset of normal and anormal cardiac data. Experimental results show that our proposed approach is reasonable and effective. The generated ECG outputs can achieve a high accuracy, and the root mean squared error (RMSE) is 0.053.