Detection of VPC using wavelet transform and fuzzy neural network

A combination of fuzzy neural network (FNN) and wavelet transform (WT) is proposed to identify the ventricular premature contraction (VPC) beats from the Holter ECG. First, the Holter ECG is transformed into different scales using a quadratic spline wavelet. Three features, including the widths of the QRS complex at scales 2/sup 3/ and 2/sup 4/ and the area of QRS at scale 2/sup 5/, are extracted and input into a four-layer FNN. The Gaussian function is adapted as the membership function in the FNN. In addition, during the training period, the backpropagation algorithm is used to train this FNN. The preliminary results indicate that the proposed method can achieve high accuracy in the detection of VPC.