Music Information Research Articles

Comparison of Lithuanian and Polish Consonant Phonemes Based on Acoustic Analysis – Preliminary Results

The goal of this research is to find a set of acoustic parameters that are related to differences between Polish and Lithuanian language consonants. In order to identify these differences, an acoustic analysis is performed, and the phoneme sounds are described as the vectors of acoustic parameters. Parameters known from the speech domain as well as those from the music information retrieval area are employed. These parameters are time- and frequency-domain descriptors. English language as an auxiliary language is used in the experiments. In the first part of the experiments, an analysis of Lithuanian and Polish language samples is carried out, features are extracted, and the most discriminating ones are determined. In the second part of the experiments, automatic classification of Lithuanian/English, Polish/English, and Lithuanian/Polish phonemes is performed.

Indian Film Songs Genre Wise Classification Based on Spectral Features

The classification of musical genres has received a lot of attention in recent years due to the several uses it can serve in music information retrieval. Additionally, since genres frequently co-occur in music tracks in Indian Film Songs, it is desirable to record and model the genre correlations in order to enhance the effectiveness of multi-label music genre classification. Present a novel multi-modal strategy that uses a music spectral feature-based classification method to address these problems. In addition, a genre correlations extraction module is offered to identify and model probable genre correlations based on the characteristics of the proposed approach.

The Effectiveness of Indian Music in Emotion Pattern Recognition under the Framework of Machine Intelligence

Experts in music therapy has suggested music as an aid to give the positive state of mind by keeping all sorts of depression and anxiety away. Music helps to bring back the original state of vibration by controlling our emotions [1]. A music is a combined effect of melody, the singer’s voice, and linguistics. The singer voice expresses the singer’s emotion like glad sorrow, anxiety, peace, tiredness and which in turns control the listener’s mental state. Indian music is analyzed and an approach for information retrieval to propose a therapeutic system through detection and identification of Indian music is initiated. Music Information Retrieval is a powerful tool to analyze different characteristics of a music. However, in this approach different traits of music are studied, and categorization has been done which leads to the therapeutic cause.

Creating musical features using multi-faceted, multi-task encoders based on transformers

Computational machine intelligence approaches have enabled a variety of music-centric technologies in support of creating, sharing and interacting with music content. A strong performance on specific downstream application tasks, such as music genre detection and music emotion recognition, is paramount to ensuring broad capabilities for computational music understanding and Music Information Retrieval. Traditional approaches have relied on supervised learning to train models to support these music-related tasks. However, such approaches require copious annotated data and still may only provide insight into one view of music—namely, that related to the specific task at hand. We present a new model for generating audio-musical features that support music understanding, leveraging self-supervision and cross-domain learning. After pre-training using masked reconstruction of musical input features using self-attention bidirectional transformers, output representations are fine-tuned using several downstream music understanding tasks. Results show that the features generated by our multi-faceted, multi-task, music transformer model, which we call M3BERT, tend to outperform other audio and music embeddings on several diverse music-related tasks, indicating the potential of self-supervised and semi-supervised learning approaches toward a more generalized and robust computational approach to modeling music. Our work can offer a starting point for many music-related modeling tasks, with potential applications in learning deep representations and enabling robust technology applications.

Emotion classification for musical data using deep learning techniques

This research is done based on the identification and thorough analyzing musical data that is extracted by the various method. This extracted information can be utilized in the deep learning algorithm to identify the emotion, based on the hidden features of the dataset. Deep learning-based convolutional neural network (CNN) and long short-term memory-gated recurrent unit (LSTM-GRU) models were developed to predict the information from the musical information. The musical dataset is extracted using the fast Fourier transform (FFT) models. The three deep learning models were developed in this work the first model was based on the information of extracted information such as zero-crossing rate, and spectral roll-off. Another model was developed on the information of Mel frequencybased cepstral coefficient (MFCC) features, the deep and wide CNN algorithm with LSTM-GRU bidirectional model was developed. The third model was developed on the extracted information from Mel-spectrographs and untied these graphs based on two-dimensional (2D) data information to the 2D CNN model alongside LSTM models. Proposed model performance on the information from Mel-spectrographs is compared on the F1 score, precision, and classification report of the models. Which shows better accuracy with improved F1 and recall values as compared with existing approaches.

Analyzing the Performance of Machine Learning Models in Music Genre Classification

Abstract: Music genre classification is a fundamental task in the field of music information retrieval (MIR) and has gained significant attention in recent years due to the rapid growth of digital music collections. This research paper presents a comprehensive review of the application of machine learning techniques for music genre classification. We explore various methodologies, feature extraction techniques, and classification algorithms used in the domain, highlighting their strengths, limitations, and recent advancements. The objective of this paper is to provide researchers and practitioners with a comprehensive understanding of the current state-of- the-art approaches, challenges, and future directions in music genre classification using machine learning.

Novel mathematical model for the classification of music and rhythmic genre using deep neural network

Music Genre Classification (MGC) is a crucial undertaking that categorizes Music Genre (MG) based on auditory information. MGC is commonly employed in the retrieval of music information. The three main stages of the proposed system are data readiness, feature mining, and categorization. To categorize MG, a new neural network was deployed. The proposed system uses features from spectrographs derived from short clips of songs as inputs to a projected scheme building to categorize songs into an appropriate MG. Extensive experiment on the GTZAN dataset, Indian Music Genre(IMG) dataset, Hindustan Music Rhythm (HMR) and Tabala Dataset show that the proposed strategy is more effective than existing methods. Indian rhythms were used to test the proposed system design. The proposed system design was compared with other existing algorithms based on time and space complexity.

MKGCN: Multi-Modal Knowledge Graph Convolutional Network for Music Recommender Systems

With the emergence of online music platforms, music recommender systems are becoming increasingly crucial in music information retrieval. Knowledge graphs (KGs) are a rich source of semantic information for entities and relations, allowing for improved modeling and analysis of entity relations to enhance recommendations. Existing research has primarily focused on the modeling and analysis of structural triples, while largely ignoring the representation and information processing capabilities of multi-modal data such as music videos and lyrics, which has hindered the improvement and user experience of music recommender systems. To address these issues, we propose a Multi-modal Knowledge Graph Convolutional Network (MKGCN) to enhance music recommendation by leveraging the multi-modal knowledge of music items and their high-order structural and semantic information. Specifically, there are three aggregators in MKGCN: the multi-modal aggregator aggregates the text, image, audio, and sentiment features of each music item in a multi-modal knowledge graph (MMKG); the user aggregator and item aggregator use graph convolutional networks to aggregate multi-hop neighboring nodes on MMKGs to model high-order representations of user preferences and music items, respectively. Finally, we utilize the aggregated embedding representations for recommendation. In training MKGCN, we adopt the ratio negative sampling strategy to generate high-quality negative samples. We construct four different-sized music MMKGs using the public dataset Last-FM and conduct extensive experiments on them. The experimental results demonstrate that MKGCN achieves significant improvements and outperforms several state-of-the-art baselines.

Deep Chnet 1-D CNN using raw audio waveform for recognizing traditional Chinese musical instrument

The classification of music information using various deep learning models is increasingly popular in the field of Music Information Retrieval research. However, as most proposed works focus on western music and musical instruments, little attention is given to traditional Chinese music. This paper proposes a 1-D Convolutional Neural Network (1-D CNN) using only raw audio waveform as input, to undertake the task of traditional Chinese musical instruments classification. This paper starts with a review of the current state of research on the related field, then discuss the proposed model and its data in detail, followed by its performance metrics and then a conclusion on the experiment. The result shows that 1-D CNN provides competitive and even superior results when compared to its 2-D versions as well as when compared to traditional models.

The Design and Development of Network Multimedia Music Teaching Based on Multiple Linear Regression Algorithm

Abstract Multimedia music information teaching is a fundamental approach to the balanced development of education in China. This paper uses the comprehensive evaluation index and multiple linear regression analysis to make statistics of network multimedia music education survey data. Then it discusses network multimedia music’s influencing factors from subjective and objective aspects. The results show a positive correlation with the final index of training, informationization funds and teaching informationization related to network multimedia music informationization. Sig. Values are all below 0.05. The three factors of information training, information fund investment and information terminal have apparent explanatory effects on the development level of information network multimedia music teaching.

Report on the 23rd International Society for Music Information Retrieval Conference (ISMIR 2022)

The annual International Society for Music Information Retrieval (ISMIR) conference is the world's leading research forum on processing, searching, organizing and accessing music-related data. The 23rd International Society for Music Information Retrieval Conference (ISMIR 2022) was held from 04--08 December, 2022 at the National Science Seminar Complex, Indian Institute of Science, Bengaluru, India. ISMIR 2022 is the first ever ISMIR conference to take place in India. Due to the changing global landscape as a result of the COVID-19 pandemic, the 23rd ISMIR conference became the first hybrid ISMIR conference, with both in-person and remote participants. The scientific program of the single-track conference comprised two keynotes, 113 research papers, six tutorials, 13 music pieces and five special sessions. The conference also included a Women in MIR (WiMIR) plenary session, industry presentations, a late-breaking/demo session, five satellite events, and multiple social events that showcased the rich musical and cultural heritage of India. In this report, we present a summary of the event, highlighting some unique challenges we faced organizing the hybrid conference and reflecting on the event for future improvements. Date: 04--08 December 2022. Website: https://ismir2022.ismir.net.

The comparison of algorithms of audio fingerprinting

With the increasing demand for song recognition on music platforms, the Audio fingerprint system is being used more and more frequently. How to improve the speed and accuracy of the Audio fingerprint system is a problem that is worth studying. Finding a better algorithm is crucial if people who use this system want to improve their efficiency of it. This paper, it is compared the accuracy and speed of extracting music information by the Multiple Hashing Algorithm and the Philips Robust Hashing Algorithm. The Multiple Hashing Algorithm can improve the accuracy of information extraction by increasing the sub-fingerprint bands, but with the increase of sub-fingerprint bands, the algorithm run time is much longer. Therefore, the Philips Robust Hashing Algorithm should be used when processing simple audio information to ensure the speed of the system. The Multiple Hashing Algorithm can improve the accuracy of information extraction by increasing the number of sub-fingerprint bands.

A study on the classification of Chinese folk instruments based on deep learning

Due to the Internet's tremendous expansion in recent years, digital music has seen explosive growth. It is becoming increasingly important to provide fast and accurate music retrieval for users, and music information retrieval has gradually become a research hotspot. Among them, the classification of musical instruments is one of the hot research directions. As an important part of the world's musical instruments, the Chinese folk musical instrument tradition has great research value, but there is less research in this direction. Traditional music classification methods can be roughly divided into two steps: manual feature extraction and traditional machine learning. But traditional music classification has several shortcomings, as follows. Firstly, manual extraction of music features has a high error rate and it is very difficult to ensure the accuracy of the features, and the labor cost is high; secondly, traditional machine learning still has problems with multiple classifications and training on large scale data. Therefore, in this paper, based on previous research, the original data features are extracted from the Mel Frequency Cepstrum Coefficient (MFCC), which is a feature that can represent musical timbre, and deep learning is proposed to be used to classify traditional Chinese folk instruments. This paper finds that the results of identifying and classifying Chinese folk musical instruments using different classical classification methods are compared when the training set and test set data are the same. According to the experimental results, the use of KNN has the best performance among these classification methods.

Applications of deep neural networks on music emotion recognition

Music Emotion Recognition (MER) is a subfield of Music Information Retrieval (MIR) that focuses on finding a relationship between music and human emotions by applying machine learning and signal processing techniques. In recent years, neural networks have achieved great success in a large number of areas, such as speech recognition and image processing, sparking many attempts to utilize neural networks in the MER task. Although new models for MER are constantly emerging, there are few systematic reviews in this field that involves the latest models and datasets. Therefore, in this paper, we provide a detailed review of this task. Our work first expounds on the practical significance and research status of the MER problem. Then, we encapsulate the background research and contributions of predecessors in both machine learning and psychology fields. Our work also includes a thorough analysis of several important datasets and their mathematical principles. Finally, we summarize four novel models from an application perspective and conclude a few potential challenges for the task.

Popularity Prediction of Music by Machine Learning Models

Each piece of music has a considerable amount of information and attributes, and each listener has their own unique taste in music, with some songs being very popular and others being relatively niche. It becomes worthwhile to examine what kinds of music are more popular. It’s clear that in this subject, the analysis and study of the available data is a "crucial first step". In this paper, we use several fitting models in machine learning as the theoretical basis, using pandas, sklearn, xgboost, and other related tools in python, to predict the popularity of music based on a dataset of music information originating from Kaggle The most suitable machine learning model is founded for predicting music popularity and the effectiveness of its fit are evaluated. This study provides a methodological basis for finding the factors influencing music popularity in post-order studies and can be a key study in determining the factors that influence the popularity of music in the future.

Music Segmentation and Similarity Estimation Applied to a Gaze-Controlled Musical Interface

Assistive technology, especially gaze-controlled, can promote accessibility, health care, well-being and inclusion for impaired people, including musical activities that can be supported by interfaces controlled using eye tracking. Also, the Internet growth has allowed access to a huge digital music database, which can contribute to a new form of music creation. In this paper, we propose the application of Music Information Retrieval techniques for music segmentation and similarity identification, aiming at the development of a new form of musical creation using an automatic process and the optimization algorithm Harmony Search to combine segments. These techniques for segmentation and similarity of segments were implemented in an assistive musical interface controlled by eye movement to support musical creation and well-being. The experimental results can be found in [https://bit.ly/2Zl7KSC].

Using multi-audio feature fusion for android malware detection

Detecting mobile Android malware is still a challenge, despite the numerous research efforts. This paper presents a static detection approach that employs music information retrieval techniques. Detection based on a single, or a few, acoustic features suffer from reduction in classification accuracy, due to the use of limited ‘views’. In this paper, we propose a multi-audio feature-fusion approach, which merges audio features of heterogeneous views in order to detect Android malware. Sixty-three standard audio signal processing features and thirty-nine biologically inspired audio features were extracted, after converting the Android application package files into waveform audio files. The biologically inspired features were derived from Mel-Frequency Cepstral Coefficients (MFCC), Gammatone Frequency Cepstral Coefficients (GFCC), and Bark Frequency Cepstral Coefficients (BFCC). Experimental results show that the proposed audio-based malware detection features are effective and need to be further studied. Using the traditional eXtreme Gradient Boosting machine learning algorithm on the CICMaldroid 2020 dataset, the proposed approach achieved accuracy, recall, f1-score, and AUC scores of 98.96%, 99.65%, 99.30% and 98.14% respectively. An average-precision (AP) score of 100% was also achieved.

Multiscale approaches to music and the embodied brain

A wide body of research is currently being devoted to investigating the multiscale processes across the brain and body, and the nature of their interactions. The purpose of this paper is to supplement these analyses of brain and body dynamics by providing a comprehensive account of the multiscale organisations also found in music, and ways in which these systems interact. We proceed in identifying scaling laws as a signature for multiscale features of a system and make the methodological choice of distinguishing 1) scale free structure from 2) scale free dynamics. We follow these distinctions in demonstrating how specifically i.) hierarchical temporal structures, ii.) long-range temporal correlations, and iii.) musical information as scale free structures relate to activity in the brain and body at various temporal or spatial scales during music perception and performance. Further, this is paralleled by fixing our focus to a limited choice of scale free dynamics, namely, iv.) resonance and entrainment, v.) power law distributions, and vi.) 1/f scaling. Though the ubiquitous scaling relations across musical activities, performance and music itself touch on the theme of universality, we outline how differing theoretical proposals concerning the relation between brain, body and musical environment arise in such contexts. Using these examples, we identify their overlap and divergence, and propose future directions for inquiry along these lines.

Theoretical Study Of Legal Protection Of Song Royalty On Digital Platforms

<p><em>The purpose of this study is to solve problems related to legal arrangements regarding song royalties or music on digital platforms that are used commercially and the issue of violations of song royalties on digital platforms today. This type of research is normative legal research with primary legal materials covering the applicable laws and regulations and secondary legal materials, including law books, the internet, and journals related to the issues studied.</em> <em>This research is descriptive, while the research material is carried out qualitatively. This research results in a conclusion that the matter of collecting royalties for songs on the Digital Platform after the issuance of Government Regulation of the Republic of Indonesia Number 56 of 2021 concerning Management of Royalties for Copyright of Songs and/or Music only targets the form of public services but does not regulate strictly in the form of Digital services that are used commercially and Legal protection regarding royalties for songs or music in Digital Platforms that are used commercially is divided into 2 (two), namely: Legal protection in the form of protection of works and rights for songwriters in the form of economic rights and moral rights, legal protection in the form of criminal sanctions and civil sanctions in the form of compensation. The responsibility of Government must immediately make derivative regulations relating to the Management of Song Royalties in Digital Platforms that are used commercially to provide protection and legal certainty for those who have economic rights/song royalties, as well as the Directorate General of Intellectual Property (DJKI) to be able to complete the Song / Music Information System (SILM) song and music data center to optimize the withdrawal and distribution of royalties to songwriters or musicians.</em><em></em></p>

A Study on Music Genre Classification using Machine Learning

Artificial Intelligence (AI) and Machine Learning can be cited as one of the greatest technological advancements in this century. They are revolutionizing the fields of computing, finance, healthcare, agriculture, music, space and tourism. Powerful models have achieved excellent performance on a myriad of complex learning tasks. One such subset of AI is audio analysis. It entails music information retrieval, music generation and music classification. Music data is one the most abstruse type of source data present, mainly because it is a tough work to extract meaningful correlating features from it. Hence a myriad of algorithms ranging from classical to hybrid neural networks have been tried on music data for a getting a good accuracy. This paper studies the various methods that can be used for music genre classification and compares between them. The accuracies we obtained on a small sample of the Free Music Archive (FMA) dataset were: 46% using Support Vector Classifier (SVC), 40% using Logistic Regression, 67% using Artificial Neural Network (ANN), 77% using Convolutional Neural Networks (CNN), 90% using Convolution-Recurrent Neural Network (CRNN), 88% using Parallel Convolution-Recurrent Neural Network (PCRNN), 73% without using Ensemble technique and 85% using Ensemble technique of Ada Boost. We defined SVC as our baseline model which had 46% accuracy, and we defined the succeeding models to achieve accuracy greater than that. ANN gave us 67% on the test dataset, which was surpassed by CNN at 77%. We noticed image based features worked better at classifying the labels than normal extracted features from the audio. A combination of CNN and RNN worked the best for the dataset, with a series CRNN model giving the best accuracy. Succeeding that, we tried to fit an ensemble model onto our dataset and analyzed its workings. This paper presents a comprehensive study of the various methods that can be used for music genre classification, with a focus on some parallel models and ensembling techniques.