There is significant evidence that music that balances emotional content and intellectual structure is fractal in nature-eg recent publications on music of Bach exhibit power law plots
this hypothesis could form the basis of an experimental regime
You can differentiate between emotions and feelings because they are two different things. The experimental aspects can be worked from the behaviors evidenced from proposed musical stimuli or contexts. Emotions in addition, are associated with esthetic patterns and other processes of expressiveness. An ethnomusicological perspective would help on the qualitative aspects starting from the cultural immersion of the realized experiments and their possible implications in the musical communication and its derived cognitive factors.
MIRToolbox is an important application in the area. The work is based in the work of Toivainen and Eerola on emotion recognition in music. Also it gives confident metrics on other acoustic features. It includes specific algorithms for music emotion recognition of a musical piece. Good luck!