In recent K-POP music, the melody line is important, but also how to design the sound that composes the song becomes an important factor. There are many various active research using artificial intelligence in the music field, such as automatically separating the desired audio from the mixture audio and composing music by themselves. In this paper, we use ‘Wave-U-Net’, which specializes in separating sound sources containing various sound sources, which are the characteristics of K-POP music or Electronic music, which are popular all over the world. The particular instrument here focuses on the ‘Bass’. We train Wave-U-Net using bass characteristics that have significantly low frequency band than other instruments. And we analyzes the bass frequency values from random songs using Mel-Spectrogram.. Lastly, we experiment to automatically analyze the chord progression of the random song. In order to make the best use of the frequency characteristics of bass instruments, experiments are conducted by comparing three methods ‘Low Pass’, ‘Bass Boost’ and ‘Low Pass + Bass Boost’ and we select the best model. Finally, based on the results of this experiments, we propose ideas that can be applied.