Abstract : Assuming two positive overlapping signals, with known shapes, the proposed method estimates the distances between their mean positions, width and area ratios. The data are two profiles representing the component shapes: no parametric model is assumed. The algorithm seeks shape equality between a linear combination of observation and first component, and the second component, in function of the area ratio. At the minimum shape difference the three parameters (distance between components, scaling factor and area ratio) are estimated. After theory, simulations are presented on Gaussian signals. Then, the method was applied on ECG signals from BSPM device during exercise on healthy people. The aim is mainly to get time distance between each T-wave and the P-wave of the following beat, on a given lead, in case of overlapping. Shape and width of the T-wave were shown to be constant before P-wave interference, which allowed taking such a real T-wave as first component model. Assumption of the same shape for the second component gave good results, as can be viewed on the reconstructed signals.