Comparing the Accuracy of CNN Model with Inception V3 for Music Instrument Recognition

Title

Comparing the Accuracy of CNN Model with Inception V3 for Music Instrument Recognition

Creator

Renju K.; Ashok Immanuel V.

Description

Identification of music instruments from an audio signal is a complex but useful task in music information retrieval. Deep Learning and traditional machine learning models are extremely very useful in many music related tasks such as music genre classification, recognizing music similarity, identifying the singer etc. Music Instrument recognition and classification would be helpful in categorizing different categories of music. Many researchers have proposed models for classifying western music instruments. But very little research has been done in identifying instruments accompanied with South Indian music. This research aims at identifying string instrument such as violin and woodwind instrument such as flute accompanied in a Carnatic music concert and also in other categories of music. In order to identify the instruments accompanied, Convolutional Neural Network model and Inception V3 models were used. The Mel Frequency Cepstral Coefficients images were extracted from the audio input and fed in to the neural network model. The model has been trained for the above mentioned instruments, tested and validated on different types of audio input. This research also evaluates the performance of Inception V3 transfer learning model with CNN model in recognizing the instruments used in different categories of music. 2024, Ismail Saritas. All rights reserved.

Source

International Journal of Intelligent Systems and Applications in Engineering, Vol-12, No. 17s, pp. 276-282.

Date

2024-01-01

Publisher

Ismail Saritas

Subject

Deep Learning; Machine Learning; Mel Frequency Cepstral Coefficients; Music Instrument recognition

Coverage

Renju K., Department of Computer Science, Research Scholar, CHRIST(Deemed to be University), Karnataka, India; Ashok Immanuel V., Department of Computer Science, Research Scholar, CHRIST(Deemed to be University), Karnataka, India

Rights

Restricted Access

Relation

ISSN: 21476799

Format

Online

Language

English

Type

Article

Identifier

https://www.scopus.com/inward/record.uri?eid=2-s2.0-85185928892&partnerID=40&md5=506577b09e00be6250351881b365a689