Sometimes using knowledge on source of signal might yield better results than just observing the data alone.

For example, if you'd know that the triangular pulse is result of a fixed-time excitation and the ramp-up arm of triangle is linear (or any other predictible shape, but that'd be more complex so let's say it's linear for sake of explanation) from the principle of measurement, you could infer the maximum peak from the angle of the ramp alone, instead of searching for the actual data point at the end of the ramp. So you could take all the points along the ramp and fit them (using a suitable method, e.g. least-square) by a straight line, which uses ALL those points to eliminate (average-out) the noise.

Jan