The process of Structural Health Monitoring aims to detect changes in material and/or geometric properties, boundary conditions or member connectivity of a structural system in time. However, the transfer of research results into engineering practice poses several challenges, especially for stiff structures with very diverse geometry that could not be well dynamically exited like low and medium-rise structures under ambient vibration conditions. Noise in modal parameters from output-only modal analysis due to variable environmental and operational conditions is considered one of the most problematic aspects of detecting structural damage using a vibration-based method. This research proposes a new way to reduce environmental noise in vibration data and dynamic parameters by merging dynamic response data from two similar shear wall buildings. The object of the study is a three-story reinforced concrete building. First, the damage features as natural frequencies and zero-order temporal moment of the vibration response are studied. Further, those feature changes are explored by means of modelling wall removal/opening introduction into the finite element models. It is established that the variation in the base excitation spectrum has more impact on dynamic response than introduced structural changes. Therefore, a time-domain feature like a zero-order temporal moment of the vibration response is not applicable for the proposed method. The appropriate damage sensitive feature vector for this approach is the difference of natural frequencies from two monitored buildings. The proposed method for fusion of vibration information from several buildings that share the same environmental and operational conditions filter out environmental noise effectively and give a clear advantage in reducing false alarm possibility during continuous and automated structural health monitoring process.