Extraction of pitch, duration and formant frequencies for emotion recognition system

Analysis of speech for recognition of emotion can be done using important characteristics such as pitch, intensity, vocal tract spectrum, duration, formant frequencies etc. Out of these characteristics, pitch, duration and formant frequencies estimation method are used for identification of angry, disgust, sad, boredom and neutral emotion identification. Analysis of above mentioned characteristics is done using TU-Berlin database (Technical University of Berlin). Pitch is determined using cepstral analysis and formant frequencies are using LPC analysis. Total duration is calculated by the extraction of vowel, semivowel, and consonant containing in that particular speech signal.