An Accurate and Fast Algorithm of Pitch Estimation Based on Relationship with Overtones

This paper introduces a fast and accurate pitch detection method. The method estimates the rough frequency of the pitch using the location of the lowest frequency peek in spectrum. Then combining Jacobsen algorithm and Fourier coefficient interpolation iterative algorithm, the fine frequency is obtained by frequency amendent. Later, according to the multiple frequency relationship between the pitch and overtone, we weighted average the frequency obtained by dividing the overtone frequency by its multiple with the frequency of the pitch peak, and take the result as the final fundamental frequency, which further improves the accuracy of the fundamental frequency. Since the frequency peek of noise is generally smaller than that of the pitch and overtone, and the overtone frequency is an integral multiple of the fundamental frequency, the interference peak caused by noise can be eliminated and the noise immunity of the algorithm can be enhanced by using these relationships. A large number of experiments show that our method not only outperforms the existing mature methods in speed and accuracy, but also maintains good robustness in noise environment.