A neural network approach to formant estimation
Xiaoyuan Zhu, L.W. Cahill · 2003
A neural network with two hidden layers is designed to estimate formant frequencies from linear predictive coding (LPC) coefficients. The formant estimation model reflects the nonlinear relationship between formants and LPC coefficients. Experiments of formant estimation on vowels and diphthongs are described and discussed. The results indicate that the neural network approach is a reliable formant estimation method.>