An Adaptive Scheme for Real-Time Detection of Patient-Specific Arrhythmias Using Single-Channel Wearable ECG Sensor

This letter proposes an adaptive time–frequ-ency (TF) scheme based on Stockwell transform (ST) for real-time detection of patient specific arrhythmias using single-channel wearable electrocardiography (ECG) sensor. A novel method, i.e., adaptive ST (AST) is proposed to improve the TF resolution by introducing three parameters to adjust the window width in classical ST. The different features are extracted from the coefficients of proposed AST to extract the nonlinear, stochastic, and complex behavior of various classes of ECG signals. The proposed AST method exhibits better TF resolution and the lateral continuity of frequency content is more continuous in AST than classical ST. These features are further classified using ensemble learning of classifiers. The proposed method is validated over Physionet data under a patient specific scheme reporting an accuracy of 97.53%. The experiments show that the proposed method can be considered as an effective TF analysis tool in distinguishing the different ECG signals in comparison of the state-of-the-art methods.