Research On Automatic Harmonization For Melody With Chinese 5-tone Modal Harmony

As one of the branches of Artificial Intelligence , Algorithmic Composition simulates the composers to harmonize the melody music. And as a cross-subject, Algorithmic Composition is worthy of being researched. Based on Chinese five-tone Modal harmonic chord theory of music theory, this dissertation describes a chorale harmonization system which uses Hidden Markov Models. To some extent , this models simulates the composers to harmonize.Firstly, we collect some chords of Chinese five-tone Modal harmonic chord and build a chord library for accompaniment, and collect more than 80 Chinese west-side folksongs with piano accompaniment used as training data set. And then, Two Hidden Markov Modals respectively for harmonizing a melody and deploying piano accompaniment figures for the melody figures are built up. In the first model, the observation value of the model is a core tone sequence in a musical piece with variable length and the hidden state is the chord label about the musical piece. In the second model, the observation value is a tone duration sequence in the musical piece and the hidden state is the piano accompaniment figures corresponding to the musical piece.when training or accompanying, we also take into account the analysis strategy for music—the struct of music. We set up a method which we use to syncopate the music into many phrases and periods. Based on this, we take the beat or bar as the auto unit to analyze the harmonic fuction and accompany the melody. At the last, in view of the music integration, based on the training result, the Viterbi Algorithm is used to generate a piano accompaniment score with nationality styles based on Chinese 5-tone Modal Harmony for an input melody .