This work presents a technique for synthesizing realistic electrocardiogram (ECG) signals by morphing two different real ECG signals. The two parent signals are interpolated using spline approximation and then over-sampled. The characteristic feature points are extracted manually from the signals and they are partitioned into component curves between the feature points. The component curves are sampled to obtain the same number of data points for each curve and the feature points are matched. One-dimensional morphing is performed to generate intermediate signals. The intermediate signals are constrained within the physiological bounds of the two real ECG signals. The realistic ECG synthesized could find application for testing an ECG-based biometric identification system and for evaluating ECG signal processing algorithms.