This research presents a music recommendation system based on multiple users' communication excitement and productivity. Evaluation is conducted on following two points. 1, Does songA recommended by the system improve the situation of dropped down communication excitement? 2, Does songB recommended by the system improve the situation of dropped down and productivity of collaborative work? The objective of this system is to recommend songs which shall improve the situation of dropped down communication excitement and productivity. Songs are characterized according to three aspects; familiarity, relaxing and BPM(Beat Per Minutes). Communication excitement is calculated from speech data obtained by an audio sensor. Productivity of collaborative brainstorming is manually calculated by the number of time-series key words during mind mapping. First experiment was music impression experiment to 118 students. Based on 1, average points of familiarity, relaxing and BPM 2, cronbach alpha factor, songA(high familiarity, high relaxing and high BPM song) and songB(high familiarity, high relaxing and low BPM) are selected. Exploratory experiment defined dropped down communication excitement and dropped down and productivity of collaborative work. Final experiment was conducted to 32 first meeting students divided into 8 groups. First 4 groups had mind mapping 1 while listening to songA, then had mind mapping 2 while listening songB. Following 4 groups had mind mapping 1 while listening to songB, then had mind mapping 2 while listening songA. Fianl experiment shows two results. Firstly, ratio of communication excitement between music listening section and whole brain storming is 1.27. Secondly, this system increases 69% of average productivity.