Reversible ECG Data Hiding: Analysis and Comparison of ANN, Regression SVM and Random Forest Regression

Abstract

The telemedicine industry is rapidly shaping the medical facilities; it starts from a basic task like booking appointments and goes up to complex tasks like diagnosis and surgery. The telemedicine industry is flourishing and a huge amount of patient data is flowing over internet. In such scenario it is very important to ensure integrity and safety of medical data. In this paper a reversible data hiding technique for ECG (electrocardiogram) signals is discussed and performance analysis of random forest (RF) regression, regression SVM (support vector machine) and artificial neural network (ANN) is done. RF, SVM, and ANN are used to predict the ECG samples, and watermark is embedded using prediction error expansion (PEE). The performances of all the models are measured using SNR (signal to noise ratio), (PRD) (percentage residual difference), & NCC (normalized cross-correlation). Models are tested for different embedding strength, ANN model has superior performance over SVM and RF. Due to reversible nature of the scheme, original signal can be completely recovered from watermarked signal.