Article Preview
Top1. Introduction
Music is not only a form of art but also a language that expresses human emotions, inner modes and affective information (Juslin, & Sloboda, 2001; Wang, Rodríguez, & Ramos, 2012; Wilson & Keil, 2001). It is generally perceived that music would not be composed, performed or comprehended without affective cognition and involvement. Because music expresses human emotions including joy, happiness, annoyance, sadness and pain, aesthetics and cognitive science recognize music as one of the powerful affective expression means.
It is recognized that music creation and appreciation are a subjective cognition process. Individuals may have different experience and understanding of the same piece of music, as well as different extends of emotionally affective effects. Therefore, it is a challenging problem to rigorously recognize and evaluate emotions of music and songs in musicology, esthetics, psychologists, and cognitive science (Juslin, & Sloboda, 2001; Wang, Rodríguez, & Ramos, 2012; Wilson, Keil, & Wilson, 2001; Hallam, Cross, & Thaut, 2008). One of the encouraging solutions for addressing the challenges is cognitive machine learning (Wang, 2015, 2016, 2017). Various machine learning algorithms are widely adapted to recognize music emotions (Yang, Lin, Su, & Chen, 2008; Bang, Kim, & Lee, 2013; Mokhsin, Rosli, Zambri, Ahmad, & Rahah, 2014; Jens, Sand, & Jan, 2015; Chin, Lin, Siahaan, Wang, & Wang, 2013).
There are two categories of methodologies for MER known as characteristic regression in the Valence-Arousal plan (Charanya, & Vijayalakshmi, 2015; Deng, Lu, & Liu, 2015; Wang, Wang, & Lanckriet, 2015; Lee, Jo, & Lee, 2011; Chin, Lin, & Siahaan, 2014; Soleymani, Aljanaki, & Yang, 2014; Soleymani, Caro, Schmidt, Sha., & Yang, 2013; Wang, Yang, Wang, & Jeng, 2012) and feature classification (Laurier, Grivolla, & Herrera, 2008; Schuller, 2010; Trohidis, Tsoumakas, Kalliris, & Vlahavas, 2008; Yang, Liu, & Chen, 2006). In this paper, music emotions were classified into four types such as pleasing, relaxing, angry and sad based on machine learning methodologies.