Abstract
The Goal of Chord Generation is to generate a suitable chord sequence for a given melody track. In this study, the best chord sequence is derived by dynamic programming, in which the state and transition probabilities are computed via music theory for Keyboard Harmonics. More importantly, we also design mathematical models using finite state machines to recognize the patterns of cadences for chord generation. Our experiments demonstrate that the introduction of cadence can increase the recognition rate of chord generation significantly. We have also compared the proposed method with those statistical methods in the literature. The advantages of the proposed method and the reason of its superiority are explained in details in this study.