A pitch tracking method mixing ACF & AMDF algorithms based on correlations
Yingjie Yang, Huanhuan Zhang, Xiue Guo · 2011
By analyzing the periodicity of autocorrelation functions (ACF) and average magnitude difference functions (AMDF), we present a pitch tracking method based on correlations of mixing ACF & AMDF algorithms. First, we calculate the normalized autocorrelation functions and normalized average magnitude difference functions, and then select a reasonable interpolation parameter from the experiments, and then last, we do linear interpolation. It effectively solves the problem that pitch period is difficult to extract, as the features are not obvious. Meanwhile, through simulation, we proved the accuracy of pitch tracking and the feasibility of implementation using this method for isolated words.