An on-chip robust Real-time Automated Non-invasive Cardiac Remote Health Monitoring Methodology

This paper introduces a novel Real-time Automated Non-invasive Cardiac Remote Health Monitoring Methodology for the detection of human condition by analyzing the ECG signal under the real-time environment. We proposed a novel System-on-Chip (SoC) architecture which has four folded modeling. After the data sensing, firstly, the classification module uses Hurst exponent as a metric in classifying the condition of the ECG signal. Secondly, if there is an abnormal detection by classifier, the Feature Extraction (FE) extracts QRS complex, P wave and T wave from ECG frames. As there is demand of ECG frames, a robust Boundary Detection (BD) mechanism is introduced to identify such frames. The fragmented-QRS (f-QRS) feature is also introduced as our third contribution to detect the fragmentation in the QRS complex and identify its morphology. The fourth step is compressing the ECG data using Hybrid compression technique for low area implementation and communicating this data to the nearest health care center by incorporating the concept of an Adaptive Rule Engine (ARE) based classifier. The proposed SoC architecture is prototyped on Xilinx Virtex 7 FPGA and it is tested on 100 patient's data from PTBDB (MIT-BIH), CSE and in house IITH DB, the percentage of accuracy obtained is 91% and it is under process for Tape-out.