4 Vowel Intensities, Amplitude Spectrum of EEG - Multiple Logistic Regression?

I have EEG recordings of the neural response to speech (a male vowel with a fundamental frequency of 100Hz). The vowel was presented to each subject (n=15) at four different intensities (55dB, 65dB, 75dB, and 85dB), and I ran FFTs and pulled the amplitude at all of the harmonics of the fundamental (H1 = 100Hz, H2=200Hz, etc.). I want to create a model that would predict which intensity was presented to the subject based on the amplitudes of the harmonics present in their respective EEGs.

Am I right to think that I should be using multiple logistic regression? I would like to implement this in R (or Matlab if that is easier somehow).

Here is some sample data for a mock-subject:

55dB 65dB 75dB 85dB
H1 0.084496 0.084538 0.078062 0.068662
H2 0.024522 0.037586 0.056271 0.058087
H3 0.0089218 0.014042 0.019386 0.024365
H4 0.0051347 0.013078 0.018572 0.014413
H5 0.0054165 0.013046 0.01866 0.02091
H6 0.0021987 0.0065719 0.0094521 0.009402

Any help or insight would be greatly appreciated!