MIDI Files Research Articles

MUSIC GENERATION USING NEURAL NETWORK

This study investigates the utilization of recurrent neural networks (RNNs) for generating music through MIDI files. By encoding musical data into sequences, RNN models are trained to learn patterns and structures inherent in compositions. Through the analysis of MIDI data and the evaluation of generated sequences, the effectiveness of RNNs in autonomously creating cohesive musical pieces is explored, advancing the frontier of AI-driven musical composition. Keywords—Music Generation, Long Short-Term Memory, Recurrent Neural Network, MIDI data .

Application of Music Retrieval & Generation

The research explores music generation through LSTM and VAEs neural network architectures, leveraging MIDI representations. LSTM specializes in sequential data processing, while VAEs compress music datasets into low-dimensional representations. Optimization of models is pursued by analyzing the relationship between training loss and epochs. Ultimately, a comparison between LSTM and VAEs determines the most effective algorithm for music generation. Traditional methods face issues like vanishing and exploding gradients, prompting exploration of deep learning approaches. The study aims to advance machine manipulation of music, facilitating new composition generation from existing MIDI files. Keywords: Music generation, LSTM, Variational Autoencoders (VAEs), MIDI representation, Sequential data, Neural network architectures, Training loss, Optimization, Deep learning, Vanishing gradient, Exploding gradient

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.

MidiBERT-Piano: Large-scale Pre-training for Symbolic Music Classification Tasks

This paper presents a benchmark study of MIDI-domain music classification using the mask language modeling approach of the Bidirectional Encoder Representations from Transformers (BERT). Specifically, with five public-domain datasets of single-track polyphonic piano MIDI files, we pre-train a 12-layer Transformer model using the BERT approach and fine-tune it for four downstream classification tasks. These include two note-level classification tasks---melody extraction and velocity prediction, and two sequence-level classification tasks---artist classification and emotion classification. Compared to strong recurrent neural network (RNN)-based baselines, the BERT approach does lead to higher classification accuracy, with less than 10 epochs of fine-tuning. Ablation studies further show that the pre-training remains effective even if the MIDI data of the downstream tasks are not seen during the pre-training stage, and that freezing the self-attention layers at the fine-tuning stage only slightly degrades the performance. For reproducibility, we share the weights of our pre-trained and fine-tuned models.

Implementation of Vector-Based Melody Extraction for Plagiarism Detection Using Szymkiewicz-Simpson Coefficient

Plagiarism is topical within the music industry. It is filled with circumstances such as the potential of massive losses coupled with a “false-positive” court ruling due to the blurred line of plagiarism factor. This research aims to solve the gray line of music plagiarism by exploring the potential of the Szymkiewicz-Simpson coefficient toward musical aspects of music. Melody and Rhythm are chosen as the main features to focus on in the research. MIDI files of music involved in court cases are used as data for the study, with limitations put on what cases can be used for the research. Using a threshold range of 0.1 to 0.25, detection accuracies for melodic plagiarism range from 45% to 60%, while rhythm plagiarism ranges from 60 to 65%. This shows that the algorithm of plagiarism detection has a tendency to detect non-plagiarism cases and is more effective towards rhythm plagiarism detection rather than melodic plagiarism detection against existing plagiarism cases.

A Study of Isan Folk Music Creation Using a Sampler for Music Education

This research represents a creative exploration in the field of music (Practice as Research: PaR) with the objectives of 1) studying the use of a sampler from Isan instruments and instruments from various cultures to create musical compositions using music software, and 2) examining the appropriate contexts for using a sampler in creative musical works. The research findings reveal that 1) Isan folk music possesses unique characteristics in terms of sound quality and performance techniques, while other musical elements can be combined with it in a technologically driven creative process. This involves connecting the concepts and methods of both traditional and diverse cultural music to create unified compositions using technology, such as recording musical notes in standard notation, exporting them as MIDI files, and then importing them into audio recording software to refine the sounds for complete songs. 2) Regarding the appropriate contexts for using a sampler in creative music, three main contexts were identified: an Isan folk music composition context, a music education context, and a commercial or music business context. In summary, the process of using a sampler for creative music works can bridge the concepts of various cultural music and integrate them with technology. Furthermore, this process can be extended to enhance interdisciplinary learning and teaching methods, making it applicable in diverse contexts.

Artistic Guidance and Emotional Expression Pathways in Piano Performance - Based on Deep Score Characterization Fusion

Abstract In the evolving domain of artificial intelligence and deep learning, the pedagogical applications of piano performance— specifically its artistic instruction and emotional expressiveness—have garnered increasing scholarly attention. This research leverages MIDI music file formats to mine necessary data, engaging in a detailed examination of the structural and characteristic elements of MIDI files. To optimize the precision of mathematical models applied in this study, MIDI music data underwent preliminary preprocessing steps, including hexadecimal encoding and feature filtering, which facilitated the transformation of MIDI features into vectors representing emotional attributes. The study introduced the development of a BiLSTM_BLS model tailored to articulate the nuances of piano performance and emotional expression. This model synthesis integrates features reflective of emotional states induced by piano playing. A mathematically derived loss function was utilized to gauge the model’s fidelity, thereby enabling an assessment of its efficacy in the realms of artistic instruction and emotional conveyance in piano performance. Findings indicate a proximity between the emotional states of Anger (0.392) and Stress (0.385), with Depressed (0.422) presenting a slightly elevated distance from these states. Moreover, the BiLSTM_BLS model achieved a performance score of 3.214 in the evaluation of artistic expression, which corresponds to 64.28 on a standardized 100-point scale. This investigation contributes to the enhancement of musical skills and theoretical knowledge among pianists, promotes cognitive development, augments emotional expression, and nurtures creativity.

A Comprehensive Review on Music Transcription

Music transcription is the process of transforming recorded sound of musical performances into symbolic representations such as sheet music or MIDI files. Extensive research and development have been carried out in the field of music transcription and technology. This comprehensive review paper surveys the diverse methodologies, techniques, and advancements that have shaped the landscape of music transcription. The paper outlines the significance of music transcription in preserving, analyzing, and disseminating musical compositions across various genres and cultures. It also provides a historical perspective by tracing the evolution of music transcription from traditional manual methods to modern automated approaches. It also highlights the challenges in transcription posed by complex singing techniques, variations in instrumentation, ambiguity in pitch, tempo changes, rhythm, and dynamics. The review also categorizes four different types of transcription techniques, frame-level, note-level, stream-level, and notation-level, discussing their strengths and limitations. It also encompasses the various research domains of music transcription from general melody extraction to vocal melody, note-level monophonic to polyphonic vocal transcription, single-instrument to multi-instrument transcription, and multi-pitch estimation. The survey further covers a broad spectrum of music transcription applications in music production and creation. It also reviews state-of-the-art open-source as well as commercial music transcription tools for pitch estimation, onset and offset detection, general melody detection, and vocal melody detection. In addition, it also encompasses the currently available python libraries that can be used for music transcription. Furthermore, the review highlights the various open-source benchmark datasets for different areas of music transcription. It also provides a wide range of references supporting the historical context, theoretical frameworks, and foundational concepts to help readers understand the background of music transcription and the context of our paper.

Multi-mmlg: a novel framework of extracting multiple main melodies from MIDI files

As an essential part of music, main melody is the cornerstone of music information retrieval. In the MIR’s sub-field of main melody extraction, the mainstream methods assume that the main melody is unique. However, the assumption cannot be established, especially for music with multiple main melodies such as symphony or music with many harmonies. Hence, the conventional methods ignore some main melodies in the music. To solve this problem, we propose a deep learning-based Multiple Main Melodies Generator (Multi-MMLG) framework that can automatically predict potential main melodies from a MIDI file. This framework consists of two stages: (1) main melody classification using a proposed MIDIXLNet model and (2) conditional prediction using a modified MuseBERT model. Experiment results suggest that the proposed MIDIXLNet model increases the accuracy of main melody classification from 89.62 to 97.37%. In addition, this model requires fewer parameters (71.8 million) than the previous state-of-art approaches. We also conduct ablation experiments on the Multi-MMLG framework. In the best-case scenario, predicting meaningful multiple main melodies for the music are achieved.

Algebraic tunings

We propose an approach to tuning systems in which octave doubling ratio is replaced by a suitable algebraic unit τ, and note frequencies are proportional to a subset of the ring . Then it is possible for many difference tones between notes in the tuning to also appear in the tuning. After outlining more general principles, we consider in detail some natural examples based on the golden ratio , limited by norm or by the number of digits in the greedy β-expansion. We discuss additive and multiplicative properties, implementation and composition using these tunings. The Online Supplement contains MIDI and websynths files to implement the tuning (based on β-expansions to ) on websynths.com and a composition Three Places.

Prototype "Kendang Jaipong" Virtual Instrument as Music Creativity Tools

Numerous organological discoveries in the form of virtual instruments have surfaced as a result of the impact of developing digital music technology. To conserve and promote music based on local culture, organological innovations in the form of virtual instruments, such as the kendang jaipong, must be implemented. The goal of this work is to describe the development of the kendang jaipong virtual instrument from organological investigations through the realization of a marketable prototype. By sampling from the Kendang Jaipong’s actual instrument and converting it into a virtual instrument based on the KONTAKT Library, the technique utilized is practice-led research, also known as practice-based research. Pre-production, production, and post-production are the three stages of the design process. This prototype makes use of 137 samples that are based on a variety of chosen kendang sounds. To enable each type of kendang sound to function as a virtual instrument in DAW and Digital Music Notation programs, the audio samples are organized in a placement system on the keyboard keys included in the KONTAKT feature. The Kendang Jaipong prototype has been successfully tested on several music productions using DAW-cast and video score. The usage of the multi-round robin's function to have a more diversified sample, as well as the addition of the midi files capability, are things that need to be developed on this prototype product so that it may be used as a machine learning by users.

Fingerprints, barcode sequences and quasi-phylogenies-Tools for analysing polyphonic music.

A data flow is presented for visualising the evolution of elementary structures of polyphonic music from early Baroque to late Romantic, using quasi-phylogenies based on fingerprint diagrams and barcode sequence data of 2-tuples of consecutive vertical pitch class sets (pcs). The present methodological study, which sees itself as a proof of concept for a data-driven approach, uses examples of music from the Baroque, the Viennese School and the Romantic era to show that such quasi-phylogenies can be generated from multi-track MIDI (v. 1) files that largely correspond to the eras and the chronology of compositions and composers. The method presented is considered to have the potential to support the analysis of a wide range of musicological questions. In the context of collaborative work on quasi-phylogenies of polyphonic music, a public data archive could be established that provides multi-track MIDI files with contextual data.

A latent rhythm complexity model for attribute-controlled drum pattern generation

Most music listeners have an intuitive understanding of the notion of rhythm complexity. Musicologists and scientists, however, have long sought objective ways to measure and model such a distinctively perceptual attribute of music. Whereas previous research has mainly focused on monophonic patterns, this article presents a novel perceptually-informed rhythm complexity measure specifically designed for polyphonic rhythms, i.e., patterns in which multiple simultaneous voices cooperate toward creating a coherent musical phrase. We focus on drum rhythms relating to the Western musical tradition and validate the proposed measure through a perceptual test where users were asked to rate the complexity of real-life drumming performances. Hence, we propose a latent vector model for rhythm complexity based on a recurrent variational autoencoder tasked with learning the complexity of input samples and embedding it along one latent dimension. Aided by an auxiliary adversarial loss term promoting disentanglement, this effectively regularizes the latent space, thus enabling explicit control over the complexity of newly generated patterns. Trained on a large corpus of MIDI files of polyphonic drum recordings, the proposed method proved capable of generating coherent and realistic samples at the desired complexity value. In our experiments, output and target complexities show a high correlation, and the latent space appears interpretable and continuously navigable. On the one hand, this model can readily contribute to a wide range of creative applications, including, for instance, assisted music composition and automatic music generation. On the other hand, it brings us one step closer toward achieving the ambitious goal of equipping machines with a human-like understanding of perceptual features of music.

The Music of Heart Rate Variability

AbstractThe variability of the heart rate reflects the regulatory status of the autonomous nervous system. For analytic purposes accessible to laypersons, the authors developed methods to transform heart rate variability into acoustical information, through either direct sound synthesis or the production of MIDI files (musical instrument digital interface) to trigger other devices. The authors describe the methods and some results and discuss applications for analytical and artistic purposes, such as music composition and biofeedback. The resulting “music” is complex and rhythmic and often has unexpected and interesting implied harmony.

Text to song synchronization using deep neural network

The most crucial aspect of the learning process is memory. There is no knowledge without memory. However, mostpeople lack memory because they memorize things through rote memorization. So in this paper, we implementeda deep neural network (DNN) model that converts plain text into a song, which is done by synchronization of aMusical Instrument Digital Surface (MIDI) file or the music with the text. MIDI files are symbolic representations ofthe music score. The melodic line of the composition is normally on one of them, while the background music ison the others. Analyzing the audio to identify voice portions where a human voice is present, as well as non-voicesegments in proximity to the voice segments, is part of a method for syncing text with audio. The text segmentsof segmented text related to the audio can be recognized and synchronized to the speech segments.

ACO and generative art – artificial music

The paper presents the adaptation process of an Ant Colony Optimization system to the task of composing music. The main aims of the study include designing and implementing the above-mentioned system, conducting research, measuring the obtained results, and drawing conclusions. The application was implemented in Python and it allows you, based on the loaded MIDI file, to generate new melodies played on various instruments. The way of composing a melody is configurable by modifying the parameters of the ACO algorithm and different composition modes. To measure the quality of the obtained results, an evaluation function was created, which is used in research to measure the correlation between the parameters of the system. Based on the conducted research, it was determined how individual parameters affect the results obtained by the system. The paper covers the presentation of theoretical knowledge, system design, implementation description, research, and conclusion.

Music Generation Using Recurrent Neural Networks

Abstract: Over the past years music has been continually evolving through the its tempos and beats as well as it’s melody. Traditionally, music is produced by a group of musicians with different instruments combining together to create a final synchronised product. In recent years, harmonies and beats were always considered to be generated manually. However, with the advent of digital technologies and software, it has become possible for machines to generate music automatically at an alarming pace. The purpose of this research is to propose a method for creating musical notes using Recurrent Neural Networks (RNN), specifically Long Short-Term Memory (LSTM)networks. To implement this algorithm, a model is created and the data is represented in the form of a percussive instrument digital interface (MIDI) file for easy access and interpretation. The process of preparing the data for input into the model isalso discussed, as well as techniques for receiving, processing, and storing MIDI files for use as input.To enhance its learning capabilities, the model should be able to remember previous details of a musical sequence and its structure. This paper discusses the use of a layered architecture in the LSTM model and how its connections interweave to create a neural network.

AN APPLICATION OF SONIFICATION AS AN ALTERNATIVE FOR THE ACCESSIBILITY OF ASTRONOMICAL IMAGES TO THE VISUALLY IMPAIRED

Astronomy's dependence on visual perception makes it difficult to be accessed by the visually impaired. Nevertheless, there are efforts to communicate astronomy to the visually impaired with the help of auditory perception and tactile perception. Screen reader applications are accessibility tools which help the visually impaired perceive images with the help of words. One of the disadvantages of screen reader is abstract astronomical concepts may be difficult to be recognized, especially for those who are visually impaired since childbirth. In this research, an image sonification program is designed as an alternative to screen reader applications to communicate astronomical images to the visually impaired using shape-approach sonification. The program was written using Python 3.8 programming language. The program is able to sonify shapes automatically and is also responsive to changes and differences in shapes. The result from program is a MIDI file which is then polished and exported using DAW to create music in the form of MP3 file. After a series of evaluation with the help of six people with visual impairment, we found that this method of sonification has potential as an alternative to the accessibility of astronomical images for the visually impaired with sufficient practice.

Long Short-Term Memory-Based Music Analysis System for Music Therapy.

Music can express people’s thoughts and emotions. Music therapy is to stimulate and hypnotize the human brain by using various forms of music activities, such as listening, singing, playing and rhythm. With the empowerment of artificial intelligence, music therapy technology has made innovative development in the whole process of “diagnosis, treatment and evaluation.” It is necessary to make use of the advantages of artificial intelligence technology to innovate music therapy methods, ensure the accuracy of treatment schemes, and provide more paths for the development of the medical field. This paper proposes an long short-term memory (LSTM)-based generation and classification algorithm for multi-voice music data. A Multi-Voice Music Generation system called MVMG based on the algorithm is developed. MVMG contains two main steps. At first, the music data are modeled to the MDPI and text sequence data by using an autoencoder model, including music features extraction and music clip representation. And then an LSTM-based music generation and classification model is developed for generating and analyzing music in specific treatment scenario. MVMG is evaluated based on the datasets collected by us: the single-melody MIDI files and the Chinese classical music dataset. The experiment shows that the highest accuracy of the autoencoder-based feature extractor can achieve 95.3%. And the average F1-score of LSTM is 95.68%, which is much higher than the DNN-based classification model.

A Deep Learning-Based Piano Music Notation Recognition Method.

In the era of rapid development of computer technology, piano music notation and electronic synthesis system can be established using computer technology, and the basic laws of music score can be analyzed from the perspective of image processing, which is of a great significance in promoting piano improvement and research and development, etc. In this paper, the Beaulieu analysis method is used to analyze the piano music notation and electronic synthesis system module. For piano sheet music, sheet music recognition is the main problem in the whole system. Through the digital recognition method, the piano sheet music feature matrix is extracted to get the piano sheet music multiplication frequency points and the envelope function needs to be extracted for better electronic synthesis of piano sheet music. The envelope function can represent the relationship between piano sound intensity and time change and finally achieve the recognition of the piano score. We extract the music information from the digital score, thus converting the music information into MIDI files, reconstructing the score, and providing an audio carrier for the score transmission. The experimental results show that the system has a correct rate of 94.4% in extracting music information from piano scores, which can meet the needs of practical applications and provide a new way for music digital libraries, music education, and music theory analysis.