This paper presents preliminary results towards an automated monitoring system for the quality assessment of insulated rail joints based on axle box acceleration (ABA) measurements. ABA signals from all the wheels of the measuring train (trailing and leading wheels, right and left rails) are processed to extract those characteristics that better represent the quality and the state of the insulated joint. We compared the responses of different insulated joints in good condition, and three cases of joints in poor condition with visible surface degradation, fastening damage and damage to the insulation layer. Different indicators are used to classify the joint condition, the most relevant being the maximum ABA response, and some frequency bands in the ABA power spectrum. Track inspections and hammer tests were performed to collect validation data of the studied insulated joints.