User Adaptive QRS Detection Based on One Target Clustering and Correlation Coefficient

This paper presents a computationally efficient user adaptive QRS detection algorithm for wearable devices that achieves an average positive prediction rate of 99.39% and sensitivity rate of 99.21% evaluated on 48 tapes of MIT-BIH arrhythmia database. The proposed user-specific QRS template avoids the complicate models and parameters used in existing algorithms while covers most situations for practical applications. The detection is based on the comparison of the correlation coefficient of the user-specific template extracted from individual user with the input ECG signal segment under detection. To reduce the computation, a novel one-target clustering is proposed to reduce the required loops from (K+1 )K/2 to 1 comparing to a K target group clustering; meanwhile, the detection judgement is triggered by the possible peaks to avoid a continuous time correlation calculation and can be taken as one-point FIR filter operation following by the dividing of the standard derivation of the input ECG signal segment.