Emotion recognition in Romanian language using LPC features

In this study we compare the recognition accuracy of the emotions when in the feature vectors are introduced the LPC coefficients, with previous results obtained with prosodic features (F0-fundamental frequency, F1-F4-formants), respectively MFCC (Mel frequency) cepstral coefficients. From the extended sets of parameters that we introduced LPCC ((linear prediction cepstral coefficients)+LPC+PARCOR (partial correlation coefficient)+LAR (log area ratio) coefficients +AC (autocorrelation coefficients), best results was obtained for PARCOR coefficients when the emotion recognition rate was around 81%.