This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License
|
||||||||
|
Paper Details
Paper Title
Content Based Audio Retrieval using Chord
Authors
  Priyanka Gaonkar,  Dr. Satishkumar Varma,  Rupali Nikhare
Abstract
Music information retrieval (MIR) is a science of retrieving information from music signal. Extracting high-level features of music such as melody, harmony or rhythm from the raw music signal is a critical and challenging process in Music Information Retrieval (MIR) systems. Using one of such feature will be helpful in searching and retrieving relevant musical audio track effectively and efficiently from large collection of musical audio tracks. Content-based concept is based on the content of a given audio document and extracting the necessary information from it.
The proposed content based audio retrieval system retrieves relevant songs from a large dataset of music audio tracks based on melody similarity. Relevant song retrieval is done by recognizing and extracting chord progressions (CPs). Chord progressions means transitions between adjacent chords which contains music information related to tonality and harmony which is helpful in effectively distinguishing whether music audio tracks are similar to each other or not. Various combination of features (4, 6 and 9) are used which consist of features like audio spectral centroid, spectral projection, audio spectrum flatness, audio spectrum spread, spectral crest factor, spectral decrease, spectral flux, spectral kurtosis, spectral Mfcc.
Then supervised statistical machine learning model such as Support Vector Machine (SVM) and Hidden Markov Model (HMM) is used for recognizing and extracting CPs from music signals. Input to the system is music audio file and output is relevant list of ordered similar audio file from the database along with their emotions like joyful, angry, depression, content or normal. Music file here belongs to any one of 5 different emotion categories. This database consists of stored predefined audio file emotions and their features vector values. There are total 300 audio files in the database out of which 70% of audio files are used for pre-storage and 30% are used for testing. Finally, based on similarity between CPs of input audio track and CPs of all songs in the database, relevant songs are returned to the user in a ranked list as the retrieval results.
Keywords- Audio retrieval, Support Vector Machine (SVM), Hidden Markov Model (HMM), Audio Features
Publication Details
Unique Identification Number - IJEDR1702249Page Number(s) - 1587-1595Pubished in - Volume 5 | Issue 2 | June 2017DOI (Digital Object Identifier) -    Publisher - IJEDR (ISSN - 2321-9939)
Cite this Article
  Priyanka Gaonkar,  Dr. Satishkumar Varma,  Rupali Nikhare,   "Content Based Audio Retrieval using Chord", International Journal of Engineering Development and Research (IJEDR), ISSN:2321-9939, Volume.5, Issue 2, pp.1587-1595, June 2017, Available at :http://www.ijedr.org/papers/IJEDR1702249.pdf
Article Preview
|
|
||||||
|