Feature Selection Approaches for Optimising Music Emotion Recognition Methods

The high feature dimensionality is a challenge in music emotion recognition. There is no common consensus on a relation between audio features and emotion. The MER system uses all available features to recognize emotion; however, this is not an optimal solution since it contains irrelevant data acting as noise. In this paper, we introduce a feature selection approach to eliminate redundant features for MER. We created a Selected Feature Set (SFS) based on the feature selection algorithm (FSA) and benchmarked it by training with two models, Support Vector Regression (SVR) and Random Forest (RF) and comparing them against with using the Complete Feature Set (CFS). The result indicates that the performance of MER has improved for both Random Forest (RF) and Support Vector Regression (SVR) models by using SFS. We found using FSA can improve performance in all scenarios, and it has potential benefits for model efficiency and stability for MER task. NTRODUCTION Music has become an indispensable part of people's lives. It plays a vital role in our world. We use music in almost every field, such as public places, entertainment, and even therapy. As the technology grows, the widespread adoption of digital audio formats, especially MP3, music distribution has become very efficient and seamless. The primary method of music consumption has shifted from retail stores to online and internet-based distribution channels. Subscription services had now become popular where the consumers now have access to much larger libraries than when albums were purchased individually. Traditional approaches to managing digital music libraries using of embedded metadata are no longer sufficient to deal with such a large database since the text cannot fully convey the expression of the musical content [1] [2], therefore the content-based music retrieval system can be ideal to handle this task more efficiency and opens a new perspective to discover music.