EmotionBox: a music-element-driven emotional music generation system using Recurrent Neural Network

Abstract

With the development of deep neural networks, automatic music composition has made great progress. Although emotional music can evoke listeners' different emotions and it is important for artistic expression, only few researches have focused on generating emotional music. This paper presents EmotionBox -an music-elementdriven emotional music generator that is capable of composing music given a specific emotion, where this model does not require a music dataset labeled with emotions. Instead, pitch histogram and note density are extracted as features that represent mode and tempo respectively to control music emotions. The subjective listening tests show that the Emotionbox has a more competitive and balanced performance in arousing a specified emotion than the emotion-label-based method.

AI-Based Affective Music Generation Systems: A Review of Methods, and Challenges

“…The RNN can be used alone [80] or in combination with a CNN [61] for generating emotion-specific affective music.…”

“…For example, Zheng and colleagues [80] have proposed a RNN-based AI-AMG system where musical features such as pitch histogram and note density are used for training the network. These features can be seen as representative of mode and tempo in music, and thus alteration of these features can lead to a change in emotional content in music.…”

“…Specific to video applications, it is also challenging for these systems to create music that can facilitate a smooth transition between different emotions. To mitigate the issue of the requirement of the large labelled dataset, researchers have tried to train the network with emotion-carrying musical features such as pitch histogram, and note density instead of training the network with a labelled musical dataset [80]. Even though the proposed method could limit the requirement related to data size to some extent, in the future researchers could aim to devise more sophisticated approaches for handling this issue along with the other challenges.…”

“…Another very important feature in affective music composition is the mode of the composition, e.g., the scale (typically major/minor) in which the music is composed, which is linked to emotional valence. A rule-based approach can be used for selecting the mode of an affective composition [65,80], where positive valence is associated with the selection of major chords, and negative valence is associated with minor chords. A combination of mode and tempo has also been used by researchers (e.g., [70,75]) for encoding different combinations of valence and arousal in an AI-AMG system.…”

Pembangkitan Musik Klasik Menggunakan Metode Long-Short Term Memory

“…Paket ini berisikan fungsi dan kelas yang memudahkan pengelolaan data MIDI dalam Python, seperti metode untuk parsing, modifikasi dan analisis file MIDI. Paket Pretty MIDI digunakan dalam berbagai permasalahan otomatisasi pembangkitan musik seperti pembangkitan musik simbolik multi-track [2], pembangkitan musik berdasarkan emosi tertentu [13], pembangkitan musik berbasis Autoencoder yang bervariasi [14] dan pembangkitan musik klasik simbolik [15]. Pemrosesan data dalam penelitian ini juga menggunakan library Pretty MIDI untuk membaca dan menampilkan notasi dalam MIDI.…”