Machine Learning Evaluation for Music Genre Classification of Audio Signals

Machine Learning Evaluation for Music Genre Classification of Audio Signals

Chetna Dabas, Aditya Agarwal, Naman Gupta, Vaibhav Jain, Siddhant Pathak
Copyright: © 2020 |Pages: 11
DOI: 10.4018/IJGHPC.2020070104
(Individual Articles)
No Current Special Offers


Music genre classification has its own popularity index in the present times. Machine learning can play an important role in the music streaming task. This research article proposes a machine learning based model for the classification of music genre. The evaluation of the proposed model is carried out while considering different music genres as in blues, metal, pop, country, classical, disco, jazz and hip-hop. Different audio features utilized in this study include MFCC (Mel Frequency Spectral Coefficients), Delta, Delta-Delta and temporal aspects for processing the data. The implementation of the proposed model has been done in the Python language. The results of the proposed model reveal an accuracy SVM accuracy of 95%. The proposed algorithm has been compared with existing algorithms and the proposed algorithm performs better than the existing ones in terms of accuracy.
Article Preview

Musical genres are created in order to describe and categorize various categories of music. They do not possess any strict definition and boundaries as they are created through cultural and marketing factors by the common public. This has led to the curiosity of some researchers to systematical study of various genres of music.

In order to fetch customers, the involuntary music analysis remains on the top list of the crucial services that the music substance providers use to sell their product. Examples include Saregama Carvaan. We have provided a framework for the design and evaluation of feature which describe musical content. For music, we have proposed various audio analysis techniques using classification and machine learning techniques. This research paper analyzes and compares the accuracies for different classifiers and suggests the best method for classifying the dataset.

The authors in the year 2002 while using machine learning techniques presented their research work on music genre classification (Tzanetakis & Cook, 2002). They also created dataset namely GTZEN ( which is widely used in the area of music genre classification since then. Research work related to features of audio and music classification has been carried out by the authors of (Mckinney et al., 2003).

Some authors also contributed to the survey and implementation work in the music genre classification arena while making use of machine learning techniques (Clark et al., 2012). An author (Sturm, 2013) has presented detailed evaluation work done in the existing literature related to genre classification. The authors of (Kumar et al., 2016) discussed a significant study based on feature extractors in concern with music genre classification. Some other significant learning related to classification seems to have been carried at various online sources (Kumar et al., 2017, Clark et al., 2012, Lidy, 2005; Bahuleyan, 2018).

Further, authors of (Li et al., 2003) specially mention that small amount of research is performed so far in relation with automating music genre classification; along with the classification accuracies supposedly reported in the existing literature are relatively low.

In (Goulart, 2012), the authors present various techniques for music genre classification incorporated with feature extraction procedure. A classification mechanism follows this stage along with exploration of parameters.

The authors of (Lidy et al., 2005) have performed a study for impressive audio feature calculation, based on the significance transformations of psycho-acoustic domain. The results of this study discovered both the problematic as well as crucial parts of the proposed algorithm.

In (Tzanetakis et al., 2002) the classification of audio signals is automatically done and musical genres with hierarchies are presented. With their proposed technique and feature sets, their research work ten musical genres are considered and a classification of 61% has been established.

The authors in a recent research work (Bahuleyan, 2018) performed performance comparison of two classes of models. In the first one, a Convolutional Neural Network model gets trained for the prediction of the associated genre label corresponding to an audio signal by utilizing its spectrograms. In the next one frequency and time domain hand designed features are utilized. The machine learning classifiers are trained using these features and their performance is compared by the authors. The experimentation is performed using the Audio data set and they their results depict an AUC component of 0.894 with the fused above-mentioned approaches.

In a research work (Finn et al., 2006), the idea of domain transfer was introduced long back where genre classifiers can be reused amongst various topics. The authors of this work also investigated various features responsible for creating classifiers for genre with transferring capabilities amongst domains of various-topics. They also showed the importance of multiple features for enhancing performance and reducing the labeling requirements for the documents.

Complete Article List

Search this Journal:
Volume 15: 2 Issues (2023)
Volume 14: 6 Issues (2022): 1 Released, 5 Forthcoming
Volume 13: 4 Issues (2021)
Volume 12: 4 Issues (2020)
Volume 11: 4 Issues (2019)
Volume 10: 4 Issues (2018)
Volume 9: 4 Issues (2017)
Volume 8: 4 Issues (2016)
Volume 7: 4 Issues (2015)
Volume 6: 4 Issues (2014)
Volume 5: 4 Issues (2013)
Volume 4: 4 Issues (2012)
Volume 3: 4 Issues (2011)
Volume 2: 4 Issues (2010)
Volume 1: 4 Issues (2009)
View Complete Journal Contents Listing