Evaluation of multipitch estimation by multi-scale product analysis
Jihen Zeremdini, Mohamed Anouar Ben Messaoud, Aïcha Bouzid · 2017
A multi-scale product comparative study for dominant pitch estimation in a mixture is presented in this paper. Indeed, the multi pitch estimation approach is based on the Autocorrelation of the Multi-scales Product of the composite signal. Then, window length, wavelet type and scales chosen for the Multi-scales Product analysis have a big influence in the multi pitch estimation. Thus, this study allows us to have the most adaptable length for the window, wavelet and scales for a better pitch estimation. We used the Cooke database and we calculated the gross pitch error rate and the root mean square error for the evaluation and the comparison.