In this paper we propose an original solution to combine the scales of multi-resolution shape descriptors. More precisely, a classifier fusion scheme is applied to shape descriptors based on ridgelets transform for symbol recognition purposes. Ridgelets coefficients are grouped into several descriptors according to their resolution and then a classifier is trained for each resolution. In this perspective, we increase the recognition rates of previous works on ridgelets-based descriptors using the GREC 2003 database.