This study proposes a methodology for constructing a music database for research purposes. We focused on the feasibility of an efficient and reliable technique to collect music stimuli that induce a variety of emotions. We selected iconic phrases from 4 famous classical pieces with 4 controls of scales, respectively. We modified each piece into 4 categories by changing the properties in terms of “mode” (converting a major piece to a minor, or vice versa) and “tempo” (changing the BPM speed faster or slower). We verified our method by a Music Information Retrieval (MIR) system. The MIR analyses showed that most of the pieces were successfully positioned in the intended categories (major/minor mode at a high/low tempo). The result suggests that this may be an efficient method to construct an objective music database that is independent of the psychological evaluations.