Formant paths tracking using methods based on Linear Prediction. Ireneusz Codello * , Wiesława Kuniszyk-Jóźkowiak Instytut Informatyki Uniwersytet Marii Curie-Skłodowskiej pl. M.Curie-Skłodowskiej 1, 20-031 Lublin, Polska Abstract In this paper we focus on formants as basic parameters for vowels recognition. We use two different algorithms for formants finding based on LP algorithm: spectral peak picking and root extraction algorithm – obtaining very good path estimations by each algorithm. We show comparison of those methods in a graphical form in our application ‘WaveBlaster’. 1. Introduction A sound is nothing but a complicated wave varying in time which is converted by our ears and brain causing a sound perception. Such a human speech sound wave can be divided into homogeneous chunks – phonemes, which are indivisible components of utterance (like letters in writing). Each phoneme has its own characteristic wave – that’s why we can distinguish one phoneme from another. If we could implement a method which recognizes phonems, then we could we could, for instance , recognize speech – we only have to find a good method for wave decomposition. The most popular method is to find all frequencies that the phonem consist of. To do so, we have to divide the wave into small frames (for instance 46ms) and compute spectrum of each frame (we use FFT or LP for this purpose). Unfortunately such a spectrum consist of to much data (frequencies), therefore we have to select only the most significant piece of information from it. One of a such significant information are ‘formants’ - local extremums of frequencies in the spectrum. Figure 1. Spectrum (in decibels) of vowel ‘a’ Smooth curve is obtain by Linear Prediction and more frequency-varying is obtained by Fourier analysis. Sampling rate is 22kHz. On the figure 1 formants have frequencies: 910Hz, 1600Hz, 3160Hz, 4400Hz – they are marked with bold crosses. Usually only four first formants are taken into account, because others are irrelevant. It is very hard to calculate formants from Fourier analysis – due to its variations. * e-mail: irek.codello@gmail.com