Multiple Fundamental Frequency Estimation Based on Harmonic Structure Model

Multiple-f0 estimation refers to the core part of music signal analysis. Although it has been considerably developed, current techniques are still not accurate and robust. Partial overlapping is one of the most difficult problem, all algorithms have troubles in processing the shared partial. In this paper, we propose a multiple-f0 estimation method based on the harmonic structure model. The candidate f0 is detected by calculating the harmonic structure matching rate and the strength, and the proportional subtraction mechanism is employed to solve the partial overlapping problem. The simulation experiments show that our algorithm is fine and computationally efficient.