Parameter optimization of awavelet-based electrocardiogram delineator with an evolutionary algorithm

A recurrent problem encountered in many algorithms proposed to detect and segment ECG waves is the adjustment of the numerous parameters used. This work presents a method to optimize these parameters with an evolutionary algorithm (EA). The signal processing chain contains a filter to remove baseline wandering, a QRS detector (Pan & Tompkins) and a wave segmentation step based on the wavelet-transform (WT). The EA adjusts the parameters of the segmentation step in order to minimize the result of a cost function which measures how close the detector is from characteristic points annotated by a cardiologist. Results obtained with the QTDB are compared with other approaches of wave segmentation for which thresholds have been experimentally defined. EAs have shown to be an effective method to solve this complex problem of multiobjective optimization