Feature Extraction and Analysis for Emotion Recognition in Songs using PRAAT Software

Emotion plays an important role in communication. Music emotion recognition is the extraction of emotion present in a song sample. Pitch, intensity and formants frequencies are important features to identify the emotion in an audio signal. In this paper, different features like pitch, intensity and formant frequencies have been extracted for emotion recognition in songs and then their analysis is carried out. PRAAT software is used to extract these features. Three different emotions namely happy, sad and angry are considered for this work. The RAVDESS (Ryerson Audio-Visual Database of Emotional Speech and Song) emotional songs database is used for emotion analysis. Finally, the significance of each feature for different emotions has been detected and analyzed.