“…The use of discrete Hidden Markov Models (HMMs) with Linear Prediction Cepstral Coefficients (LPCC), Log Frequency Power Coefficients (LFPC), total signal energy (E), teager energy (TE), fundamental frequency (F0) and values of the formants (FF) has reached 72% of recognition rate (Nedeljković, Ðurović, 2015). In the case of Support Vector Machine (SVM) approach, the results oscillated between 62.78% and 91.3% depending on which test setup was used (Hassan, Damper, 2010;Milošević et al, 2016). The results obtained so far on the Polish database PES have been contrasted: 50.73% using k Nearest Neighbours (kNN) and Mel Frequency Cepstral Coefficients (MFCC) (Kamińska et al, 2013), whereas phoneme level formant features combined with Binary Decision Trees (BDT) give 81.9% (Ślot et al, 2009).…”