ECG Compression Using Machine Learning Technique with Wavelet Transform

ECG (Electrocardiogram) is the graphical representation of heart activities. As the number of heart patients increases day by day and no definite solution is found except death data. The data storage demands a large space to acquire all the data of the patients. The storage can be optimal when the signal is compressed. Through different techniques that have been approached Machine Learning technology for ECG data compression is explored a little. The authors of this article have used the machine learning technique. as a Neural Network model to compress the ECG signal. However, the model is modified with weight modification. The initial weights are considered wavelet coefficients and are further modified through hidden layers. The use of wavelet transform helps us to suppress the noise and decompose it with relevant information. To train the model, a back propagation learning technique is used. The training accuracy is 99.2 and the PRD (Percent Root Mean Square Difference) and CR (Compression Ratio) of SVDB (Supraventricular Arrhythmia Database) are 98.2 and 1.24 and for MIT-BIH arrhythmia it is 97.72 with 1.19 which is shown in the result section.