Music Genre Classification Using Machine Learning

International Journal of Trend in Scientific Research and Development (IJTSRD)
Volume 5 Issue 4, May-June 2021 Available Online: www.ijtsrd.com e-ISSN: 2456 – 6470
Music Genre Classification using Machine Learning

Seethal V1, Dr. A. Vijayakumar2
1Master of Computer Application, 2Professor,
1,2Department of Computer Application, Jain Deemed-to-be University, Bangalore, Karnataka, India
ABSTRACT How to cite this paper: Seethal V | Dr. A.

Music genre classification has been a toughest task in the area of music Vijayakumar "Music Genre Classification
information retrieval (MIR). Classification of genre can be important to clarify using Machine
some genuine fascinating issues, such as, making songs references, discovering Learning" Published
related songs, finding societies who will like that particular song. The in International
inspiration behind the research is to find the appropriate machine learning Journal of Trend in
algorithm that predict the genres of music utilizing k-nearest neighbor (k-NN) Scientific Research
and Support Vector Machine (SVM). GTZAN dataset is the frequently used and Development
dataset for the classification music genre. The Mel Frequency cepstral (ijtsrd), ISSN: 2456- IJTSRD41263
coefficients (MFCC) is utilized to extricate features for the dataset. From 6470, Volume-5 |
results we found that k-NN classifier gave more exact results compared to Issue-4, June 2021, pp.829-831, URL:
support vector machine classifier. If the training data is bigger than number of www.ijtsrd.com/papers/ijtsrd41263.pdf
features, k-NN gives better outcomes than SVM. SVM can only identify limited
set of patterns. KNN classifier is more powerful for the classification of music Copyright © 2021 by author (s) and
genre. International Journal of Trend in Scientific
Research and Development Journal. This
KEYWORDS: Machine Learning, Mel Frequency Cepstral Coefficients, k-Nearest is an Open Access article distributed
Neighbors, Support Vector Machine under the terms of
the Creative
Commons Attribution
License (CC BY 4.0)
(http://creativecommons.org/licenses/by/4.0)
I. INTRODUCTION RELATED WORKS

A music genre is a conventional category that recognizes a Music genre classification has been a hardest task in the field
couple of pieces of music as having a spot with a typical of MIR. Tzanetakis and Cook spearheaded the work on music
practice or set of shows. It is to be recognized from musical genre classification utilizing ML techniques. They made the
design and musical style, yet eventually these terms are a GTZAN dataset as a norm for genre classification.
portion of the time used equally. The innovative thought of
Nilesh M. Patil.[2] have depicted an automated classification
music infers that these classifications are consistently
framework model for music genres. It first discovered good
theoretical and debatable, and a couple of genres may cover.
features for each music genre. To get feature for the
With the appearance of MP3 and other audio coding classifiers from the dataset, features like MFCC, chroma
schemes, music content investigation has become a frequencies, spectral centroid, zero-crossing rate were
developing exploration region. Genres gives an extremely utilized. K-NN and SVM classifiers were arranged and used to
valuable description of a musical piece[12]. The advent of characterize, each yielding shifting degrees of exactness in
huge music collections has represented the test of how to prediction.
recover, browse, and suggest their contained items. One
Sergio Oramas.[3] have shown an way to deal with consider
approach to facilitate the access of huge music classification
and combine multimodal data representations for genre
is to keep label explanations of all music resources. Labels
classification is proposed. Analyses on single and multi-label
can be added either manually or automatically. However,
genre classification are completed, evaluating the impact of
because of the high human effort needed for manual labels,
the distinctive learned representations and their
the execution of automatic labels is more cost effective.
combinations. Results on the two trials show how the
However, an absence of label consistency exists in the music
aggregation of learned representations from various
community. This prompts challenges in comparing the
modalities improves the exactness of the classification.
performance of genre classification algorithms across
databases[3]. Muhammad Asim Ali [4]had is covered ML algorithm that
predict the genres of songs utilizing k-NN and SVM. This
In this paper we compared K-NN and SVM classifiers to
paper additionally presents similar examination between k-
characterize the following genres: Classical, Jazz, Metal, Pop,
NN and SVM with dimensionality return and afterward
Blues, Country, Disco, Hip-jump, Metal, Reggae, Rock. GTZAN
without dimensionality decrease by means of principal
dataset is the recommended dataset for the music genre
component analysis (PCA). From results they found that
classification. It comprises 1000 sound records each having
both k-NN and SVM gave more exact outcomes contrast with
30 seconds term. We utilized MFCC to remove data from our
the outcomes with dimensionality reduction.
information.
@ IJTSRD | Unique Paper ID – IJTSRD41263 | Volume – 5 | Issue – 4 | May-June 2021 Page 829
International Journal of Trend in Scientific Research and Development (IJTSRD) @ www.ijtsrd.com eISSN: 2456-6470
II. PROPOSED SYSTEM is 30 seconds long. Ten classifications are available in this
The advent of huge music collections has represented the dataset containing 100 tracks each. Each track is in .wav
test of how to recover, browse, and suggest their contained format. It contains sound records of 10 genres [2].
items. One approach to facilitate the access of huge music
B. Mel Frequency Cepstral Coefficients (MFCC)
classification is to keep label explanations of all music
The first phase in any music genre classification system is to
resources. Labels can be added either manually or
extract features i.e. recognize the segments of the audio
automatically. However, because of the high human effort
signal that are useful for distinguishing the linguistic content
needed for manual labels, the execution of automatic labels
and disposing of the wide range of various stuff which
is more cost-effective. To solve this issue, we compared K-
conveys data like background noise, emotion etc. [2].
NN and SVM classifiers to classify the genres by using the
GTZAN dataset. From the results, we found that the k-NN
gave more accurate outcomes compared to the SVM
classifier [3].
Figure 3 Steps in MFCC

 We partition the signal into a few short frames to keep a
n audio signal consistent.
Figure 1 Block Diagram  At that point we attempt to recognize various

frequencies present in each frame
The first step for genre classification is to separate features
and segments from the sound records. It includes  Push the power spectra into the Mel filter bank and
recognizing the linguistic contents and disposing of noise. assemble the energy in each filter to entire it. With this
Pitch and MFCC features are removed from sound records. we get the energy existing in the diverse frequency
These features are utilized to traina KNN classifier. At that regions. Equation for Mel scale is:
point, new sound records that should be arranged go M (f) = 1125 ln(1+f/700 )
through a similar feature extraction. The trained KNN  We calculate the logarithm of the filter bank energies.
classifier predicts which one of the 10 genres is the nearest This empowers people to have features closer to what
match [8]. they can hear.
 Calculating the Discrete Cosine Transform (DCT) of the
outcome décor relates the filter bank energies with one
another.
 We first keep first 13 DCT coefficients disposing of the
greater DCT coefficients that can present errors by
addressing changes in the filter bank energies[2].
C. k-Nearest Neighbors Algorithm (k-NN)
The principle machine learning technique we utilized was
the k-nearest neighbors (k-NN) as it is extremely well-
known for its simplicity of execution. K-NN algorithm can be
utilized for both classification and regression [9]. These are
the two properties that would characterize KNN -
 Lazy Learning algorithm- KNN is a Lazy Learning
algorithm, it doesn't have a specific training stage and
uses all the information for training while classification.
 Non-parametric learning algorithm- KNN is a non-
parametric learning algorithm, it doesn't expect
anything about the secret data[9].
Working of k-NN Algorithm
K-Nearest Neighbors (KNN) Algorithm utilizes feature
similarity to predict the values of new data points which
further suggests that the new data point will be allocated a
Figure 2 Work Flow value dependent on how eagerly it facilitates with the points
A. Data Gathering in the training set[11].
MARSYAS is an open source WWW for sound handling with  For executing any algorithm, we need dataset. So, during
specific supplement on sound information data users. For the initial step of KNN, we should stack the training
music genre classification GTZAN dataset utilized which has similarly as test data.
a collection of 1000 sound records. Every one of the records
International Journal of Trend in Scientific Research and Development (IJTSRD) @ www.ijtsrd.com eISSN: 2456-6470
 After that, we have to take the value of K i.e. the nearest IV. CONCLUSION
data points. K can be any number. We have presented a methodology for automatically
extracting musical features from audio files and classifying
 Calculate the distance between test data and each row of
them according to their genre in this paper. We feature
training data with the assistance of Euclidean distance.
extraction and selection, and finally classification process.
To calculate Euclidean distance is.
KNN is a supervised learning classifier that is easy to
implement. From results we found that k-Nearest Neighbors
classifier gave more accurate outcomes compared to support
vector machine classifier. If training data is much larger than
 Now, based on the distance value, sort them in number of features, k-NN gives better results than SVM. SVM
ascending order. can only identify limited set of patterns. Overall KNN is more
effective classifier which gave 75% accuracy for the
 Next, it will pick the top K rows from the arranged array. classification of music genre.
 Now, it will allocate a class to the test point dependent V. REFERENCES
on most frequent class of these rows[11]. [1] Wing W. Y. Ng, Weijie Zeng, Ting Wang(2020) “Multi-
D. Support vector Machine Algorithm (SVM) Level Local Feature Coding Fusion for Music Genre
Support Vector Machine is a model for classification and Recognition”.
regression. It can deal with linear and non-linear issues and [2] Nilesh M. Patil, Dr. Milind U. Nemade(2017) “Music
function admirably for some reasonable issues. The Genre Classification Using MFCC, K-NN and SVM
algorithm makes a hyperplane which isolates the data into Classifier”.
classes[6]. The target of the SVM algorithm is used to settle
the decision boundary that can seclude n-dimensional space [3] Sergio Oramas, Francesco Barbieri, Oriol Nieto, Xavier
into classes so it can put the new data point in the right Serra (2018) “Multimodal Deep Learning for Music
category. This decision boundary is known as a hyperplane. Genre Classification”.
SVM picks the vectors that assistance in making the [4] Muhammad AsimAli, Zain Ahmed Siddiqui (2017)
hyperplane. These extreme cases are called as support “Automatic Music Genres Classification using Machine
vectors, and algorithm is named as Support Vector Learning”
Machine[6].
[5] Ahmet Elbir; Hilmi Bilal Çam; Mehmet Emre
III. RESULTS AND DISCUSSIONS Iyican; BerkayÖztürk; Nizamettin Aydin(2018) “Music
Accuracy of classification by different genres with k-NN and Genre Classification and Recommendation by Using
SVM algorithms is varied. The achievement rate of k-NN was Machine Learning Techniques”.
75% yet the blues genre was misconstrued as rock. The k-NN
did badly while perceiving blues with a recognizing [6] Anirudh Ghildiyal; Komal Singh; Sachin Sharma
percentage of 66%. The SVM misconceived classical genre as (2020) “Music Genre Classification using Machine
jazz or hip-hop, yet the rock genre was precisely related to Learning”.
progress rate of 94%. Similarly, the k-NN did well with [7] KonstantinMarkov; Tomoko Matsui(2014) “Music
perceiving entire categories however then again it likewise Genre and Emotion Recognition Using Gaussian
erroneously distinguished disco with rock and metal with Processes”.
jazz. The k-NN classification had the success rate of country
was 73% however with rock genre it was simply 40%. Hip [8] Ali Karatana; Oktay Yildiz(2017) “Music genre
hop genre had the achievement rate of 75%. The classical classification with machine learning techniques”.
and raggae had success rate of 69% which is better than SVM [9] DivyanshRana(2019) “Music Recommendation System
however trouble to differentiates with many other genres. Using machine learning algorithms”.
By and large we found that k-NN is more successful classifier
which gave 70% of accuracy. [10] J. Umamaheswari; A. Akila(2019) “ An Enhanced
Human Speech Emotion Recognition Using Hybrid of
Genres K-NN classifier SVM classifier PRNN and KNN”.
Classical 0.69 0.62
[11] QhansaDi’Ayu Putri Bayu; Suyanto Suyanto;
Country 0.73 0.70
AndityaArifianto (2019) “Hierarchical SVM-kNN to
Pop 0.70 0.65
Classify Music Emotion”.
Hip-Hop 0.75 0.55
Blues 0.66 0.80 [12] JeremyReed, Chin-Hui Lee ”A Study on Music Genre
Reggae 0.69 0.14 Classification Based on Universal Acoustic Models”.
Accuracy 75% 65%

Music Genre Classification Using Machine Learning

Uploaded by

Copyright:

Available Formats

Music Genre Classification Using Machine Learning

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Music Genre Classification Using Machine Learning

Uploaded by

Copyright:

Available Formats

International Journal of Trend in Scientific Research and Development (IJTSRD)

Music Genre Classification using Machine Learning

ABSTRACT How to cite this paper: Seethal V | Dr. A.

I. INTRODUCTION RELATED WORKS

Figure 3 Steps in MFCC

Figure 1 Block Diagram  At that point we attempt to recognize various

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.