Approximate Logarithmic Squarers for QRS Identification on Arrhythmia Detection and Classification

Reducing power consumption in battery-powered medical edge devices is highly required. This study aims to enhancing energy efficiency in the Pan-Tompkins algorithm used for QRS identification by adopting approximate computing techniques. This paper specifically focuses on the squaring step within the algorithm, investigating approximate logarithmic squarers to improve arrhythmia classification and detection accuracy. We evaluate power and area reduction rates, as well as accuracy of QRS identification and arrhythmia classification and detection. The best candidate achieves a 96.76% power reduction and an 89.45% area reduction while maintaining a 98.13% arrhythmia classification and detection accuracy.