Sei sulla pagina 1di 4

CHAPTER 1

INTRODUCTION
1.1 Background
Music can be produced from a single musical instrument to a combination of
certain musical instruments. In order to be able to play a song with an instrument,
people simply need to find sheet music or music score. Chord is a sound
produced from the set of three or more notes which usually included in the music
score. The sound, which is produced from chord, has a harmonic quality. Chords
have many variations such as major chords, and minor chords. Some musical
instruments that can be used with chord are guitar and piano.
It is not easy to distinguish or recognize chords in the song. Chord
recognition usually performed by highly trained musicians. The reason is that
their hearing ability has been trained very well to distinguish and recognize
chords. On the other hand, for a beginner it is difficult to recognize and
distinguish the chord. Therefore, this research develops a type of system where
computer has the capability to extract chords from a sound file. In another word,
tries to make computer doing something more than human can do.
Other than that, chord is also needed to harmonize melody, and in order to
play the song, an understanding in learning the chords is needed. Therefore, a
chord recognition research is something that is interesting to be developed.
Chord recognition can be achieved with help from artificial intelligence.
Chord recognition and artificial intelligence have an interesting relationship. A
neural network is used as the base in deciding on which chord is used. An
artificial neural network is an information-processing system that has certain
performance characteristics in common with biological neural networks (Fausett,
1994: 1). Artificial Neural Network ability can solve problems like human does.
It can learn a pattern to recognize similar patterns. This gives the learning ability
for Artificial Neural Network in solving problems. Artificial Neural Network also
has the ability to be more fault-tolerance. With a better configuration, neural
network can give a nice result, although the data that is trained contains a similar
value.

Besides that, this is also based on the previous research that had been carried
out by previous music scientists. Osmalskyj, et al., in 2012 recognize only 10
chords, and able to recognize several instruments such as piano, and violin by
creating its dataset and using feed-forward neural network.
Upon this, an Artificial Neural Network will be used as the method to find the
closest match to the real chord that is being played.

1.2 Problem Formulation


The problem definitions from this thesis are:
1. How to extract frequency from wav file?
2. How Neural Network work in analyzing chord samples?
3. How musical chords can be recognized especially guitar and piano?
4. What is a good neural network design to recognize chord?
1.3 Hypothesis
A neural network design that is able to recognize single chord accurately
should be able to recognize chords of instrumental song accurately. Basically,
this is because instrumental song consists only of many chords. Moreover, A
configuration with small learning rate is able to give a better result for
recognizing chord.

1.4 Scope
In this report, system boundaries toward each part of variables which
are observed for system design and development. This system can recognize
piano or guitar as instrument and recognize the chord. This report focuses on
design boundary, without eliminating its final purposes as followed:
1. This research focuses on chord identification of music files especially
minor and major chords.
2. The extension of the song file is .wav, sample rate 22050 sample/
second (22, 5 kHz) and mono channel.
3. This research focuses on chord identification of acoustic guitar and
piano without any vocals.
4. This research recognize the chord from one instrument at a time, not
simultaneously from 2 or more instruments

5. Using backpropagation neural network.


6. Using learning rate and number of hidden nodes as parameter.
7. Neural Network uses only 1 hidden layer.

1.5 Aims and Benefits


Aims for this research project are:
1. To be able to recognize piano chords and guitar chords from audio file.
2. To design a good Backpropagation Neural Network architecture for
analyzing chord.
Benefits from the implementation of this research project are:
1.

Can be used as reference for future research such as song chord


analyzer.

2.

User is able to extract a chord from instrumental song file.

1.6 Research Methodology


Methodology that is used to collect data and generate information to
enhance this research are:
a. Index Review
With performing literature review towards Fourier transform as sound
and frequency analysis techniques. Literature reviews for backpropagation
neural network and music theories are implemented as foundation to
recognize chord based on the frequencies.

Evaluation of the previous

researches is implemented too.


b. Design and Implementation System
Design the algorithm to analyze frequency of the sound. Determine
guitar and piano chord by using neural network algorithm. The application is
built using Java programming language.
c. Evaluation of the Application
Evaluating the application after Design and Implementation system is
crucial to test the functionality and the performance. This is done to ensure
the quality assurance of the application and algorithm.

1.7 Structures
Below are 5 chapters of this report:

Chapter 1 Introduction
In this chapter discussing about problem background, system
boundary, purpose and benefit of this report, methodology that is being used
and writing systematic of the report.

Chapter 2 Theoretical Modelling


This chapter gives explanation about theoretical basis which support
in the making of the system such as musical theory, Fast Fourier Transform,
Pitch Class Profile, and Neural Network Backpropagation.

Chapter 3 System Analysis and Design


This chapter discusses about overall system flowchart, pseudocode
and system interface design.

Chapter 4 Implementation and Evaluation


This chapter describes about implementation alongside with
subjective evaluation of the system and objective evaluation of the
algorithm.

Chapter 5 Conclusion
This chapter explains the conclusion that can be deduced after the
implementation of this system, and testing. Also contains feedbacks that
will be used for future research.

Potrebbero piacerti anche