Arrhythmia is a heart disease where the heart rate is inconsistent. For some arrhythmias that can cause sudden cardiac arrest, the patient needs to be sent to the hospital for immediate treatment. Most of the current electrocardiogram (ECG) devices are bulky, cost expensive, and does not include the self-classification or interpretation ability. Hence it is not suitable for small clinics and patients to use as the first screening devices. This paper proposed a SoC-based implementation of arrhythmia detector by using embedded software design. It able to analyze the heart rate variability (HRV), diagnose and classify the arrhythmias in terms of ventricular fibrillation (VF), premature ventricular contraction (PVC), and two degree heart block (2 ^o Block) base on the R-R interval properties. The ECG signal is pre-processed and extract the R peak using Pan and Tompkins algorithm. The arrhythmia can be detected based on knowledge-based classification of the R-R intervals. The proposed system is prototyped on the Altera Video and Embedded Evaluation Kit with Multi-Touch (VEEK-MT) FPGA development board. Results shows that the proposed system able to classify the aforementioned arrhythmia types with convincing average detection accuracy in range of 85.71% to 93.61% based on MIT-BIH database.