The design of meaningful audio features is a key need to advance the state-of-the-art in Music Emotion Recognition (MER). This work presents a survey on the existing emotionally-relevant computational audio features, supported by the music psychology literature on the relations between eight musical dimensions (melody, harmony, rhythm, dynamics, tone color, expressivity, texture and form) and specific emotions. Based on this review, current gaps and needs are identified and strategies for future research on feature engineering for MER are proposed, namely ideas for computational audio features that capture elements of musical form, texture and expressivity that should be further researched. Finally, although the focus of this article is on classical feature engineering methodologies (based on handcrafted features), perspectives on deep learning-based approaches are discussed.

This work advances the music emotion recognition state-of-the-art by proposing novel emotionally-relevant audio features. We reviewed the existing audio features implemented in well-known frameworks and their relationships with the eight commonly defined musical concepts. This knowledge helped uncover musical concepts lacking computational extractors, to which we propose algorithms - namely related with musical texture and expressive techniques. To evaluate our work, we created a public dataset of 900 audio clips, with subjective annotations following Russell's emotion quadrants. The existent audio features (baseline) and the proposed features (novel) were tested using 20 repetitions of 10-fold cross-validation. Adding the proposed features improved the F1-score to 76.4% (by 9%), when compared to a similar number of baseline-only features. Moreover, analysing the features relevance and results uncovered interesting relations, namely the weight of specific features and musical concepts to each emotion quadrant, and warrant promising new directions for future research in the field of music emotion recognition, interactive media, and novel music interfaces.

This research addresses the role of lyrics in the music emotion recognition process. Our approach is based on several state of the art features complemented by novel stylistic, structural and semantic features. To evaluate our approach, we created a ground truth dataset containing 180 song lyrics, according to Russell's emotion model. We conduct four types of experiments: regression and classification by quadrant, arousal and valence categories. Comparing to the state of the art features (ngrams - baseline), adding other features, including novel features, improved the F-measure from 69.9, 82.7 and 85.6 percent to 80.1, 88.3 and 90 percent, respectively for the three classification experiments. To study the relation between features and emotions (quadrants) we performed experiments to identify the best features that allow to describe and discriminate each quadrant. To further validate these experiments, we built a validation set comprising 771 lyrics extracted from the AllMusic platform, having achieved 73.6 percent F-measure in the classification by quadrants. We also conducted experiments to identify interpretable rules that show the relation between features and emotions and the relation among features. Regarding regression, results show that, comparing to similar studies for audio, we achieve a similar performance for arousal and a much better performance for valence.

Classical machine learning techniques have dominated Music Emotion Recognition. However, improvements have slowed down due to the complex and time-consuming task of handcrafting new emotionally relevant audio features. Deep learning methods have recently gained popularity in the field because of their ability to automatically learn relevant features from spectral representations of songs, eliminating such necessity. Nonetheless, there are limitations, such as the need for large amounts of quality labeled data, a common problem in MER research. To understand the effectiveness of these techniques, a comparison study using various classical machine learning and deep learning methods was conducted. The results showed that using an ensemble of a Dense Neural Network and a Convolutional Neural Network architecture resulted in a state-of-the-art 80.20% F1 score, an improvement of around 5% considering the best baseline results, concluding that future research should take advantage of both paradigms, that is, combining handcrafted features with feature learning.

This research addresses the role of lyrics in the music emotion recognition process. Our approach is based on several state of the art features complemented by novel stylistic, structural and semantic features. To evaluate our approach, we created a ground truth dataset containing 180 song lyrics, according to Russell’s emotion model. We conduct four types of experiments: regression and classification by quadrant, arousal and valence categories. Comparing to the state of the art features (ngrams - baseline), adding other features, including novel features, improved the F-measure from 68.2%, 79.6% and 84.2% to 77.1%, 86.3% and 89.2%, respectively for the three classification experiments. To study the relation between features and emotions (quadrants) we performed experiments to identify the best features that allow to describe and discriminate between arousal hemispheres and valence meridians. To further validate these experiments, we built a validation set comprising 771 lyrics extracted from the AllMusic platform, having achieved 73.6% F- measure in the classification by quadrants. Regarding regression, results show that, comparing to similar studies for audio, we achieve a similar performance for arousal and a much better performance for valence.

The Music Emotion Recognition (MER) field has seen steady developments in recent years, with contributions from feature engineering, machine learning, and deep learning. The landscape has also shifted from audio-centric systems to bimodal ensembles that combine audio and lyrics. However, a lack of public, sizable and quality-controlled bimodal databases has hampered the development and improvement of bimodal audio-lyrics systems. This article proposes three new audio, lyrics, and bimodal MER research datasets, collectively referred to as MERGE (for Music Emotion Recognition next GEneration, the research project that funded its creation), which were created using a semi-automatic approach. To the best of our knowledge, this is the largest bimodal quality-controlled static MER dataset. Furthermore, the dataset is annotated both with discrete emotion categories (according to Russell's four emotion quadrants) and continuous arousal-valence values. To comprehensively assess the proposed dataset and establish a baseline for benchmarking, we conducted several experiments for each modality, using feature engineering, machine learning, and deep learning methodologies. Additionally, we propose and validate fixed train-validation-test splits. The obtained results confirm the viability of the proposed datasets, achieving the best overall result of 81.74% F1-score for bimodal classification.