1. I
ABSTRACT
REpeating Pattern Extraction Technique (REPET), a novel and simple approach for
separating the repeating “background” from the non-repeating “foreground” in a mixture.
The basic idea is to identify periodically repeating patterns in the audio (e.g., a guitar riff or
a drum loop), and then separate the repeating “background” from the non-repeating
“foreground” (typically the vocal line). This is embodied in an algorithm called Repeating
Pattern Extraction Technique (REPET). In this project the performance of the repeating
algorithm using the Mel Frequency Cepstral Coefficients (MFCC) with the help of
MATLAB is implemented.
.
2. II
LIST OF FIGURES
FIGURE NO. FIGURE NAME
PAGE
NO.
3.1 Method to Estimate the MFCC Components 12
3.2 Various Components of MFCC 13
3.3 Improved MFCC Iterations 14
3.4
Comparative Analysis of only Repet Algorithm and Repet +
MFCC
14
3.5 Overview of the REPET algorithm 15
3.6
Representation of Mel-Frequency Cepstrum Coefficients
(MFCC) calculation
16
3.7 Mel scale filter bank 17
4.1 Matlab Window 19
5.1
Block Diagram of the Overall Design Analysis of Music &
Voice Segregation
20
5.2 Process to create MFCC features 21
5.3 Voice Algorithm Flow Chart 22
4. IV
LIST OF EQUATIONS
EQUATION NO. EQUATION NAME PAGE NO.
3.1 Windowing Signal 16
3.2 Conversion of time domain into frequency domain 16
3.3 Energy in a frame 17
5.1 Calculation of the beat spectrum 25
5.2 Calculation of the repeating segment model 26
5.3 Calculation of the repeating spectrogram model 26
5.4 Calculation of the soft time-frequency mask 27
5. V
List of Publications
Sr.
No.
Name of
Authors
Paper Title
Name of Conference
and Journal
Status of
Paper
1
Snigdha
Bhattacharjee,
Ms.Amruta
Moon
Overview of
Performance
Enhancement of Repet
Algorithm Using
MFCC
International Journal of
Scientific Engineering
and Research,
ISSN 2347‐3878
Published
2
Snigdha
Bhattacharjee,
Ms.Amruta
Moon
Performing
Enhancement of Repet
Algorithm from
Foreground &
Background Sound
Separation Using
MFCC
IEEE Sponsored
International Journal and
Conference of Computer,
Information Technology
and Bioinformatics,
ISSN: 2278-7593
Published
and
Presented
