Melody Generation Research Articles

Comfortable sound design based on auditory masking with chord progression and melody generations using automatic composition

Unpleasant noise is a crucial problem, that we often face in daily life. One of such a problem is in the case of dental treatment. A dental instrumental noise that includes peaky components in the high-frequency band makes doctors and patients discomfort. To reduce discomfort, we proposed a method to mask unpleasant noise with a comfortable sound. The proposed method was automatically generating a comfortable cord progression. In previous subjective test results, it shows that degree of discomfort reduces. The remaining problem is that degree of comfort improves. This is due to the fact that the previous method utilizes the musical element of chord, but does not introduce the musical elements of melody and rhythm. Therefore, in this paper, we propose a new method to introduce the melody generation with automatic composition (AI composition) into the previous method based on the automatic generation of the chord progression. The proposed method automatically generates a chord progression to follow the peak frequency of dental treatment sounds and to add a melody following the chord progression with AI composition assistance tool to improve comfort and to design comfortable sound.

생성형 인공지능(Generative AI)을 활용한 대중음악 창작교육 방안 연구 - ChatGPT 활용방안을 중심으로

The education curriculum for music composition utilizing generative AI, particularly ChatGPT, aligns with the new direction for improving creative and integrative abilities. This program is based on the enhanced capabilities in the musical realm brought by the upgrade to GPT 4.0, thereby allowing for personalized music creation support across various music genres and styles. The introduction phase focuses on learning to use AI and acquiring basic knowledge of popular music. The development phase involves undertaking creative tasks such as lyric, chord, and melody generation. In the conclusion phase, participants reflect on the process and share what they have learned. This study is significant as it serves as a model that provides concrete examples of music creation activities.

Melody Generation using Deep Learning: Unleashing the Power of RNN and LSTM

This project aims to develop a novel approach for piano melody generation using Recurrent Neural Networks (RNN) and Long Short-Term Memory (LSTM) models in deep learning.The suggested models will be trained on a dataset of MIDI files with piano melodies to use sequential learning capabilities and capture the complex patterns and relationships present in musical compositions. [1] The project aims to gen- erate a variety of melodies that are both musically coherent and diverse by experimenting with various network designs, hyperparameters, and training procedures. The developed tunes will be evaluated primarily on their originality, conformity to stylistic elements, and general quality. The results of this study could lead to new developments in AI-driven music composition as well as opportunities for computational creativity in the music industry.

Sofcomputing approach to melody generation based on harmonic analysis

This work aims to create an ANN-based system for a musical improviser. An artificial improviser of "hearing" music will create a melody. The data supplied to the improviser is MIDItype musical data. This is the harmonic-rhythmic course, the background for improvisation, and the previously made melody notes. The harmonic run is fed into the system as the currently ongoing chord and the time to the next chord, while the supplied few dozen notes performed earlier will indirectly carry information about the entire run and the musical context and style. Improvisation training is carried out to check ANN as a correctlooking musical improvisation device. The improviser generates several hundred notes to be substituted for a looped rhythmicharmonic waveform and examined for quality.

Hierarchical multi-head attention LSTM for polyphonic symbolic melody generation

Creating symbolic melodies with machine learning is challenging because it requires an understanding of musical structure and the handling of inter-dependencies and long-term dependencies. Learning the relationship between events that occur far apart in time in music poses a considerable challenge for machine learning models. Another notable feature of music is that notes must account for several inter-dependencies, including melodic, harmonic, and rhythmic aspects. Baseline methods, such as RNNs, LSTMs, and GRUs, often struggle to capture these dependencies, resulting in the generation of musically incoherent or repetitive melodies. As such, in this study, a hierarchical multi-head attention LSTM model is proposed for creating polyphonic symbolic melodies. This enables our model to generate more complex and expressive melodies than previous methods, while still being musically coherent. The model allows learning of long-term dependencies at different levels of abstraction, while retaining the ability to form inter-dependencies. The study has been conducted on two major symbolic music datasets, MAESTRO and Classical-Music MIDI, which feature musical content encoded on MIDI. The artistic nature of music poses a challenge to evaluating the generated content and qualitative analysis are often not enough. Thus, human listening tests are conducted to strengthen the evaluation. Qualitative analysis conducted on the generated melodies shows significantly improved loss scores on MSE over baseline methods, and is able to generate melodies that were both musically coherent and expressive. The listening tests conducted using Likert-scale support the qualitative results and provide better statistical scores over baseline methods.

Melody generation based on deep ensemble learning using varying temporal context length

Melody generation based on deep ensemble learning using varying temporal context length

EmoMusicTV: Emotion-Conditioned Symbolic Music Generation With Hierarchical Transformer VAE

Emotion is one of the most crucial attributes of music. However, due to the scarcity of emotional music datasets, emotion-conditioned symbolic music generation using deep learning techniques has not been investigated in depth. In particular, no study explores conditional music generation with the guidance of emotion, and few studies adopt time-varying emotional conditions. To address these issues, first, we endow three public lead sheet datasets with fine-grained emotions by automatically computing the valence labels from the chord progressions. Second, we propose a novel and effective encoder-decoder architecture named EmoMusicTV to explore the impact of emotional conditions on multiple music generation tasks and to capture the rich variability of musical sequences. EmoMusicTV is a transformer-based variational autoencoder (VAE) that contains a hierarchical latent variable structure to model holistic properties of the music segments and short-term variations within bars. The piece-level and bar-level emotional labels are embedded in their corresponding latent spaces to guide music generation. Third, we pretrain EmoMusicTV with the lead sheet continuation task to further improve its performance on conditional melody or harmony generation. Experimental results demonstrate that EmoMusicTV outperforms previous methods on three tasks, i.e., melody harmonization, melody generation given harmony, and lead sheet generation. Ablation studies verify the significant roles of emotional conditions and hierarchical latent variable structure on conditional music generation. Human listening shows that the lead sheets generated by EmoMusicTV are closer to the ground truth (GT) and perform slightly worse than the GT in conveying emotional polarity.

Expressing and Developing Melodic Phrases in Gamelan Skeletal Melody Generation Using Genetic Algorithm

Expressing and Developing Melodic Phrases in Gamelan Skeletal Melody Generation Using Genetic Algorithm

Changes in Music Curriculum Education in Colleges and Universities Based on Cloud Classroom Assistive Technology

Abstract With the rapid development of science and technology, cloud classroom teaching has gradually become a trend. The cloud-based music teaching aid training system designed in this paper learns to map rhythmic and pitch features through music sequence generation technology. The onset detection algorithm of spectral flux detects the onset of the note. The Bezier curve is introduced into the melodic line feature generation algorithm to construct a polynomial function for interpolation between the first and the last endpoints, and the emotion is utilized to guide fractal music creation. Finally, the effectiveness of this auxiliary teaching system in college music teaching has been verified. The results show that the optimal Hamming distance of the system is 0.353 in the music melody generation experiment, which indicates that the system can quickly generate melody lines for students in music teaching and improve their learning efficiency. The system can significantly improve the rationality and efficiency of music skill training, which is worth promoting.

Research on multi-system interactive pop music composition based on artificial intelligence technology

Abstract The purpose of this paper is to analyze MIDI technology for pop music composition and the application of composition in artificial intelligence. Based on deep learning, a SeqVAE-based music melody generation algorithm is designed, the possibility of a pulse neural network-based music melody generation algorithm is explored, and a multi-system interactive pop music composition system is established and tested. This system's accuracy rate for bass and treble recognition was tested to be 4%, 4.1%, 2.47%, and 2.66, respectively, compared to the traditional recognition rate. The recommendation satisfaction rate was 96 when the number of queries was 50, the response speed increase time was less and kept within 3 seconds, and the CPU occupancy rate was within 40%.

Using deep learning and genetic algorithms for melody generation and optimization in music

Using deep learning and genetic algorithms for melody generation and optimization in music

An automatic music generation and evaluation method based on transfer learning.

In recent years, deep learning has seen remarkable progress in many fields, especially with many excellent pre-training models emerged in Natural Language Processing(NLP). However, these pre-training models can not be used directly in music generation tasks due to the different representations between music symbols and text. Compared with the traditional presentation method of music melody that only includes the pitch relationship between single notes, the text-like representation method proposed in this paper contains more melody information, including pitch, rhythm and pauses, which expresses the melody in a form similar to text and makes it possible to use existing pre-training models in symbolic melody generation. In this paper, based on the generative pre-training-2(GPT-2) text generation model and transfer learning we propose MT-GPT-2(music textual GPT-2) model that is used in music melody generation. Then, a symbolic music evaluation method(MEM) is proposed through the combination of mathematical statistics, music theory knowledge and signal processing methods, which is more objective than the manual evaluation method. Based on this evaluation method and music theories, the music generation model in this paper are compared with other models (such as long short-term memory (LSTM) model,Leak-GAN model and Music SketchNet). The results show that the melody generated by the proposed model is closer to real music.

MelodyDiffusion: Chord-Conditioned Melody Generation Using a Transformer-Based Diffusion Model

Artificial intelligence, particularly machine learning, has begun to permeate various real-world applications and is continually being explored in automatic music generation. The approaches to music generation can be broadly divided into two categories: rule-based and data-driven methods. Rule-based approaches rely on substantial prior knowledge and may struggle to handle large datasets, whereas data-driven approaches can solve these problems and have become increasingly popular. However, data-driven approaches still face challenges such as the difficulty of considering long-distance dependencies when handling discrete-sequence data and convergence during model training. Although the diffusion model has been introduced as a generative model to solve the convergence problem in generative adversarial networks, it has not yet been applied to discrete-sequence data. This paper proposes a transformer-based diffusion model known as MelodyDiffusion to handle discrete musical data and realize chord-conditioned melody generation. MelodyDiffusion replaces the U-nets used in traditional diffusion models with transformers to consider the long-distance dependencies using attention and parallel mechanisms. Moreover, a transformer-based encoder is designed to extract contextual information from chords as a condition to guide melody generation. MelodyDiffusion can automatically generate diverse melodies based on the provided chords in practical applications. The evaluation experiments, in which Hits@k was used as a metric to evaluate the restored melodies, demonstrate that the large-scale version of MelodyDiffusion achieves an accuracy of 72.41% (k = 1).

Melody Generation from Lyrics with Local Interpretability

Melody generation aims to learn the distribution of real melodies to generate new melodies conditioned on lyrics, which has been a very interesting topic in the area of artificial intelligence and music. However, a challenging issue still limits the quality and reliability of melody generation conditioned on lyrics: how to enhance the interpretability between the input lyrics and generated melodies so humans can understand their relationships. To solve this issue, in this article, we propose a model for melody generation from lyrics with local interpretability, which contains two significant contributions: (i) Mutual information between input lyrics and generated melody is exploited to instruct the training of the network, which avoids the loss of content consistency during the training stage. (ii) Transformer is explored to efficiently extract semantic features from lyrics sequences, which provides more interpretable correlations between different syllables in lyrics. Experiments on a large-scale dataset with paired lyrics-melodies demonstrate that the proposed approach can generate higher-quality melodies from lyrics compared with existing methods.

Generating Music With Emotions

AbstractWe focus on the music generation conditional on human emotions, specifically the positive and negative emotions. There is no existing large-scale music datasets with the annotation of human emotion labels. It is thus not intuitive how to generate music conditioned on emotion labels. In this paper, we propose an annotation-free method to build a new dataset where each sample is a triplet of lyric, melody and emotion label (without requiring any labours). Specifically, we first train the automated emotion recognition model using the BERT (pre-trained on GoEmotions dataset) on Edmonds Dance dataset. We use it to automatically ‘`label’' the music with the emotion labels recognized from the lyrics. We then train the encoder-decoder based model to generate emotional music on that dataset, and call our overall method as Emotional Lyric and Melody Generator (ELMG). The framework of ELMG is consisted of three modules: 1) an encoder-decoder model trained end-to-end to generate lyric and melody; 2) a music emotion classifier trained on labeled data (our proposed dataset); and 3) a modified beam search algorithm that guides the music generation process by incorporating the music emotion classifier. We conduct objective and subjective evaluations on the generated music pieces, and our results show that ELMG is capable of generating tuneful lyric and melody with specified human emotions.

Conditional hybrid GAN for melody generation from lyrics

Conditional hybrid GAN for melody generation from lyrics

(Retracted) Algorithm of generating music melody based on single-exposure high dynamic range digital image using convolutional neural network

Music is a very important art. However, professional music production has a relatively high investment cost and a single model, which cannot meet people’s individual needs. Single-exposure high dynamic range (HDR) digital images based on convolutional neural networks (CNNs) can solve this problem. A comparative experiment was carried out. The final experimental results show that in terms of the quality of melody generation, the algorithm is almost the same as manual creation. In terms of generation speed, the CNN single-exposure HDR digital image model is 1.1 times faster than the traditional algorithm model, which proves its advantage of higher creative ability. Popularizing it in the music creation market at the current stage can effectively improve the efficiency of music creation and promote the intelligent development of the music market.

How can generative adversarial networks impact computer generated art? Insights from poetry to melody conversion

Recent advances in deep learning and generative adversarial networks (GANs), in particular, has enabled interesting applications including photorealistic image generation, image translation, and automatic caption generation. This has opened up possibilities for many cross-domain applications in computer generated arts and literature. Although there are existing software-based approaches for generating musical accompaniment of a given poetry, there are no existing implementation using GANs. This work proposes a novel poetry to melody generation conditioned on poem emotion using GANs. A dataset containing pairs of poetry and melody based on three emotion categories is introduced. Furthermore, various GAN architectures including SpecGAN and WaveGAN were explored for automatic melody synthesis for a given class of poetry. Conditional SpecGAN produced the best melodies according to quantitative metrics. Melodies produced by SpecGAN were evaluated by volunteers who deemed the quality to be above average.

MeloSol Corpus

This data report introduces the MeloSol corpus, a collection of 783 Western, tonal monophonic melodies. I first begin by describing the overall structure of the corpus, then proceed to detail its contents as they would be helpful for researchers working in the field of computational musicology or music psychology. In order to contextualize the MeloSol corpus in relation to other corpora in the literature, I present descriptive statistics of the MeloSol corpus alongside the The Densmore Collection of Native American Song and The Essen Folk Song Collection. I suggest possible future uses of this corpus including extending research investigating Western tonality, perceptual experiments needing novel ecological stimuli, or work involving the musical generation of monophonic melodies in the style of Western tonal music.

WITHDRAWN: Music Composition Feasibility using a Quality Classification Model based on Artificial Intelligence

There are several explanations for building music systems, including computer-generated settings, playing musical instruments, reflecting on the art of sound recordings, and others. The main challenge in music composition identifies the student's time, and it shows dissatisfaction with their lack of basic musical skills. The music compositions help develop a student's greater stronger vocabulary to determine that they are rising musicians. Therefore, this paper suggests the Music Composition and Melody Generation Adjustable (MCMGA) structure develop adaptive music. MCMGA's main objective is to build music in real-time that reflects various moods that can accomplish through a single combination of the chord-sequence generator on a cross graph. A search melody predictor and an emotional-expression concept have been analysed. Melody generation uses a novel technique for evolution that combines the Feasible Population Method (FPM) with vector optimization. FPM helps to explore a Pareto front of various solutions under vector optimization functions that are creatively equivalent. To test the program, two quantitative user studies are carried out. The first is based on music generation and the second on valence expression by adding dissonances. The study's outcome shows that every other part of the production and propagation increases the perceptual quality of the generated music and the relative effectiveness of the interpretation of reactivity through dissonance based on these parameters music experience 96.7%, usability and accessibility 98.4%, interaction analysis 99.5%, music teaching analysis 96.3%, music performance anxiety analysis 29.8% Stage fright analysis 22.5%. • A search melody predictor and an emotional-expression concept have been analysed. • To test the program, two quantitative user studies are carried out. • Vector optimization is simply interpreted and self-explaining of music algorithms.