A Model of Music Perceptual Theory Based on Markov Chains

Music perceptions can be regarded as the expectancies for which the brain processes temporal statistics to predict future rhythms. The different patterns of the expectancy streams represent the different styles of music. In this paper, we present a model for music style recognition based on a music cognitive theory combined with machine learning approach. First, we establish a Markov chain with eight states where each state represents a certain composition mode derived from the Implication-Realization (IR) theory. Then we use a clustering method to detect music styles hidden in compositions. The results are identical to the conclusions in musicology, which confirms the effectiveness of our method.