It appears that what they did was to take a 2-D image and assign the vertical axis to frequency (the top is high frequencies and the bottom is low frequencies) and assign the horizontal axis to time. Since points of light appear isolated, that translates into a single frequency for each star. When a large glowing cloud is encountered, that translates to pink noise, like a "hiss". If you watch it carefully you can actually predict the appearance of certain tones and hisses just by understanding how their translation is working. Of course the assignments made are totally arbitrary. If the picture were rotated 90 degrees the entire piece would sound different.