Abstract
The first part of this paper describes an algorithm for estimating the fundamental frequency F0 of a speech signal, using an autocorrelation algorithm. After that, it was shown that, due to the discrete structure of the autocorrelation function, the accuracy of the fundamental frequency estimate largely depends on the sampling period TS. Then, in order to increase the accuracy of the estimation, an interpolation of the correlation function is performed. Interpolation is performed using a one parameter (1P) Keys interpolation kernel. The second part of the paper presents an experiment in which the optimization of the 1P Keys kernel parameter was performed. The experiment was performed on test Sine and Speech signals, in the conditions of ambient disturbances (N8 Babble noise, SNR = 5 to -10 dB). MSE was used as a measure of the accuracy of the fundamental frequency estimate. Kernel parameter optimization was performed on the basis of the MSE minimum. The results are presented graphically and tabularly. Finally, a comparative analysis of the results was performed. Based on the comparative analysis, the window function, in which the smallest estimation error was achieved for all ambient noise conditions, was chosen.