IRJET-Virtual Music Guide for Beginners using MATLAB and DSP Kit

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 05 Issue: 06 | June-2018 www.irjet.net p-ISSN: 2395-0072
© 2018, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 2360
Virtual Music Guide For Beginners Using MATLAB and DSP Kit
Amey B. Bhat1, Ashish W. Lonkar2, Ojas S. Bhargave3, Dr. Virendra Shete4
1,2,3Students, Dept. of Electronics and Telecommunication Engg, MIT College of Engg, Pune, Maharashtra, India
4Head of the Department, Dept. of Electronics and Telecommunication Engg, MIT College of Engg, Pune,
Maharashtra, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - The paper proposes asystemwhichaimsatguiding
the music enthusiasts in learning the basics of Indian classical
music through instrument playing. Generally, a physical
instructor is required for teaching the instruments to a
beginner; this project aims at eliminating thisneedand makes
the process hassle free and independent of guide. The basic
idea of project is that it exploits the fact that each Raag is
associated with certain frequency and this frequency is
processed further, to achieve the objective. It also notifies the
user if he goes wrong according to the laws of Indian Classical
Music in his practice sessions. The musical notes played by the
user are nothing but the frequencies or the pitch to be
extracted and processed. As the project involves intensive
audio processing, DSP (Digital SignalProcessing)Kitis usedas
hardware component, while for software assistance MATLAB
is used.
Key Words: Raag, Frequency, Pitch detection, Audio
Processing, Digital Signal Processing.
1. INTRODUCTION
Indian Classical Music is based on melody rather than
harmony. The basis of melody lies in the numerous Ragas
available in Indian Classical Music. It takes many years for a
student of music to understand and distinguish between
different Ragas. Understanding the nature of all Ragas is
difficult. It requirescontinuousanddedicatedeffortforavery
long period to pick up this ability. Moreover, this skill
depends very much on the musicalsense of a person. Whena
Raga is sung by a musician, itis appreciatedatdifferentlevels
by different listeners. Musical Sound has two identifying
characteristics; loudness and pitch. Loudness is power, as it
depends on the amplitude or the intensity of the
corresponding wave, and is measured in decibels. The pitch
of a musical sound is determined mainly by itsfrequencyand
is a measureof how "high" or "low" a toneis,andismeasured
in hertz (Hz). The notes of the same pitch from a Guitar,Sitar,
Violin and a Flute are entirely different in quality and are
instantly recognizable.
1.1 Objectives
A] We are accepting two inputs from the user:
1. Base Frequency [scale]: This is the frequency of the
reference note of that particular tune. For instance,inIndian
classical Music, the note ‘Sa’ will be considered as the
reference note. So if the scale is ‘C#’ or ‘Kaali 1’ then ‘Sa’ is
equal to the frequency of ‘C#’ or ‘Kaali 1’ which is 552.6 Hz
for the fifth octave. Rests of the notes in the same tune are
then calculated according to this base reference frequency.
This reference note is called as the scale of the particular
tune.
2. RAAG: This is a concept in Indian Classical Music which
is nothing but a defined set of musical notes which are used
in that particular tune. So the RAAG is that set of musical
notes or frequencies.
B] Objectives of our system:
1. Extracting the frequency from the music using signal
processing with the help of DSP. This will be used for tuning
purpose.
2. Determining and indicating whether the tuneplayed by
learner is correct or not according to the “RAAG” specified
earlier.
1.2 Concept Of SwarAndSaptakInIndianClassicalMusic
Indian Classical Music has seven swaras, with their
identities as below:- Sa, Re, Ga, Ma, Pa, Dha, Ni. These seven
swaras are further sub-divided as ‘Teevra’ and ‘Komal’. Sa
and Pa are ‘AchalSwaras’, which means that they can’t be
sub-divided into Teevra and Komalswaras. Thus, there are
12 swaras that include Shuddha, Teevra and Komalswaras.
Now, to understand the phenomenon of ‘Saptak’, we shall
use harmonium keyboard or the synthesizer as showninFig
-1. Here, both the Swaras and Saptaks could be seen in the
figure.
Fig -1: Concept of Swaras and Saptaks
2. LITERATURE SURVEY
Lyudmila Sukhostat, YadigarImamverdiyev,[1]intheirwork
described about thevariouspitchdetectionalgorithmswhich
are required for audio processing. The major types of the
algorithms are using time domain characteristics, frequency
domain characteristics, and last one being the hybrid

characteristics (both time domain and frequency domain).
For timedomain analysis,zero crossingsandautocorrelation
measurements areimportant.Theimportantalgorithminthe
time domain analysis is the Average Magnitude Difference
Function (AMDF), which has low computational cost and
relatively easy to implement. The algorithm implemented
here, in this work, is used below in our design. Also, various
algorithms were tested under noisy conditions, and the best
algorithm was found out.
Sanjivani S. Bhabad, et al. [2] worked on a project which
utilized various techniques on pitchdetectionintimedomain
and frequency domain. Here, algorithm for autocorrelation
was discussed in detail. Secondly, the Average Magnitude
Difference Function (AMDF)algorithm was implemented.As
specified in this project, the advantage of using AMDF is that
it requires only the subtraction of the signals. This project
was based on speaker recognition and the results for pitch
calculation showed that the pitch value for different user for
same vowel is different.
From the book written by Rulph Chassaing and Donald Reay,
[3] westudied the basic of the TMS320C6713DSKalongwith
its Code Composer Studio (CCS). CCS provides Integrated
Development Environment (IDE), for real time digital signal
processing, written in the C Language. The version of CCS
provided in this book was V3.3, while in our design we have
used V5.5. Installation and other external files requiredwere
also referred. The introductory part of this book gives all the
details of the board with its size, processor usedandexplains
about the AIC23 Codec. The AIC23 Codec performs the
functions of Analog-to-Digital Conversion (ADC) as well as
Digital-to-Analog Conversion (DAC). The next part of the
book gives the details about the architecture of the
TMS320C6713 DSK, giving the details about the functional
block diagram of the processor. The next concept which we
referred was about the DSP/BIOS and RTDX using MATLAB
and C Language. DSP/BIOS (Built in Operating System),
provides real timeanalysis and data transfer forapplications
running on TI DSP. There are various threads/ functions in
the DSP/BIOS, which have already pre-defined priority. For
instance, the HardwareInterrupts(HWIs),havingthehighest
priorities. Real Time Data Exchange (RTDX) [4] offers
developers continuous bi-directional data exchange in real
timewith minimal variation on theapplication.Also,itiseasy
to program on both target and host, thus making it flexible
with a variety ofapplications.Finally, MATLAB support tools
were referred from variety of examples explained in this
book.
The paper published by Savitha S Upadhya, [5] described
about time domain pitch detection methods using
Autocorrelation and AverageMagnitude DifferenceFunction
(AMDF). This paper gives detailed analyses on the
computation of the AutocorrelationFunction(ACF).Here,the
positions of the peaks were found out and distancesbetween
the successive peaks were calculated. Also, the comparison
between voices and unvoiced signals were done, and the
speaker was identified.Thoughtherecognitionofthespeaker
was not a part of our project, but the technique used forpitch
detection was same one.
3. SYSTEM BLOCK DIAGRAM AND THE COMPONENTS
Here, we are going to use Digital Signal Processing Kit i.e.,
TMS320C6713. The input will be taken through Lavalier
Microphone and will be given to this kit. The audio signal
received will be processed by this processor, and the results
will be displayed on the PC terminal. The DSP Kit will require
a power supply of +5 Volts. This could be seen in Fig -2.
Fig -2: Hardware block diagram of proposed system
3.1 Components
1. TMS320C6713 (DSP): TheC6713DSK,asshowninFig-3,
is a low-cost standalone development platform that enables
users to evaluate and develop applications for the TI C67xx
DSP family. The DSK also serves as a hardware reference
design for the TMS320C6713 DSP. Schematics, logic
equations and application notes are available to ease
hardware development and reduce time to market. The
specifications could be seen from its datasheet. [6]
Fig -3: Board diagram of TMS320C6713 DSK
2. MICROPHONE: A microphone is a transducer that
converts sound into an electrical signal. Several different
types of microphone are in use, which employ different
methods to convert the air pressure variations of a sound
wave toan electrical signal. The audiosignal can be captured

using a microphone. A lavalier microphone, as shown in Fig -
4, (also known as a lav, body mic, collar mic, neck mic) is a
small electret or Ribbon diaphragm used for television,
theatre,and public speakingapplicationsinordertoallowfor
hands-free operation. They are most commonly provided
with smallclips forattachingto collars,ties,orotherclothing.
The term referred to any small microphone that could be
hooked into the buttonhole of the lapel of a coat. The lapel
microphoneoffered freedom of movement. Itsspecifications
are given in Table-1.
Fig -4: Lavalier Microphone
Sr.
No.
Parameter Values
1. Brand Any
2. Type Wired In-Ear
3. Audio Jack 3.5 mm
4. Frequency
Response
20 - 20000 Hz ± 3 dB
5. Sensitivity 92 – 110 dB
6. Power Handling &
Impedance
32Ω : 100mW peak /
050mW rated
16Ω : 200mW peak /
100mW rated
7. Driver Types Dual (Bass & Mid + High)
8. Sound Pressure
Level (SPL)
85-120 dB SPL/mW
9. Total Harmonic
Distortion (THD)
< 1%
Table -1: Generic Earphone Specifications
3.2 Software
1.MATLABR2014b:MATLAB (matrix laboratory)isa multi-
paradigm numerical computing environment. A proprietary
programming language developed by Math Works, MATLAB
allows matrix manipulations, plotting of functions and data,
implementationof algorithms,creationof userinterfaces,and
interfacing with programs written in other languages,
including C, C++, C#, Java, Fortran and Python. MATLAB
R2014b introduces a new MATLAB graphics system. New
default colors, fonts, and styles make your data easier to
interpret.
2. CODE COMPOSER STUDIO - INTEGRATED
DEVELOPMENT ENVIRONMENT: Code Composer Studio is
an integrated development environment (IDE)thatsupports
TI's Microcontroller and Embedded Processors portfolio.
Code Composer Studio comprises a suite of tools used to
develop and debug embedded applications. It includes an
optimizing C/C++ compiler,source codeeditor, project build
environment, debugger, profiler, and many other features.
The intuitive IDE provides a single user interface taking you
through each step of the application development flow. Code
Composer Studio combines the advantages of the Eclipse
software framework with advanced embedded debug
capabilities from TI resulting in a compelling feature-rich
development environment for embedded developers.
4. BLOCK DIAGRAM
The following block diagram Fig -5, gives us the whole flow
of our system. In other words it is the step by step working
and analysis of our design. The major blocks would include
Audio signal, Frequency extraction, Raag authenticationand
at the end the displaying content in the form of musical
notes.
Fig -5: Block diagram / Flow of proposed system
 The input sound is taken through the ‘Audio SignalInput’,
which is nothing but the microphone in our design.
 This output is then given to the ‘Filter and Noise
Reduction Block’, which consists of ‘Low Pass Filter’, in
order to remove noise and other disturbances.
 The ‘Frequency (Pitch) extraction’ block is used to take
out the frequencies from the clean audio signal (sound).
 The ‘RAAG Authenticator’ would consist of laws and
regulations ofIndianClassicalMusic.Adatabaseofmusic
would be stored in this block, and would compare with
the music and thereby indicating whether it is correct or
not.
 The ‘Input {Keys}’ blockwouldbeinterfacingofKeypad.It
would have options to select the base frequency and the

type of ‘RAAG’ which we will select after recording and
then compare.
 The block ‘Realization in a musical way’ is used, as we
have to display the Sur/Notes(Sa,Re,Ga,etc)ratherthan
the musical frequency. This objective would be achieved
by coding. For instance, if the tune played is at 440Hz,
rather than displaying 440Hz, we would display ‘Sa’ on
our device.
 These Sur/Notes will be then displayed on the PC
Terminal and it will also indicate whether the tune
played is correct or not.
5. ALGORITHMS USED
The various pitch detection methods are:
Existing pitch detection methods can be divided into three
groups:
 The first primarily uses the time-domain characteristics.
 The second uses the properties of frequency domain.
 The hybrid methodscombinepropertiesofbothtime-and
frequency-domain characteristics.
Time-domain methods are performed directlyonthespeech
signal. For this type of pitch detectors peaks, valleys,
frequency of zero-crossings, and autocorrelation
measurements are very important. Time-domain
measurements will provide a good evaluation of
fundamental frequency if a quasi-periodic signal is
appropriately processed to minimize the effects of the
formant structure. Difficulties of time-domain methods for
the male voice with low-pitch tone occur due to the analysis
of fixed length frames at 30–40 milliseconds which,asa rule,
is not suitable for fundamental frequency evaluation.
Frequency-domain pitch detectors use the property that if
the signal is periodic in the time domain, then the frequency
spectrum of the signal will consist of a series of impulses at
the fundamental frequency and itsharmonics. Suchmethods
include Harmonic Product Spectrum (HPS), cepstrum pitch
determination, and Linear Predictive Coding (LPC).
5.1 Autocorrelation:
Related to the time domain feature detector is the
autocorrelation method. The autocorrelation of the signal is
first formed:
And for discrete signals:
The main peak in the autocorrelation function is at the zero
lag location (m = 0). The location of the next peak gives an
estimate of the period, and the height gives an indication of
the periodicity of the signal. The method usually requires a
number of periods of data to form a reliable estimate,
and thus some averaging of the frequency signal is
unavoidable. The method often exhibit difficulty in
detecting the period of a periodic signal which ismissingthe
fundamental harmonic in the harmonic series. Periodic but
pathological signals can be devised to cause nearly anypitch
detection algorithm to fail.
5.2 Average Magnitude Difference Function (AMDF):
The pitch detector/tracker is a refinement of the Average
Magnitude Difference Function (AMDF) detectors, the
earliest of which is that of Miller and Weibel. Methods ofthis
type have also been called comb-filter methods. The AMDF
pitch detector forms a function which is the compliment of
the autocorrelation function, in that it measures the
difference between the waveform and a lagged version of
itself.
The generalized AMDF function is:
and fundamental frequency is the smallest period value
taken as:
For discrete signals AMDF function is given by:
The quantity k is set to 1 for average magnitude difference,
and other values for other related methods. The zero lag
(m=0) position of the AMDF function is identically zero, and
the next significant null is a likely estimate of the period.
Other nulls will occur at integer multiplesoftheperiods.The
signal is preprocessed to aid in detection ofthefirst null.The
difficulties of using this pitch detection method arise from

the issues of finite sampling rate, noise, and signal
stationary. If the signal is truly periodic with period Tu,
and Tu is an integer multiple of the sampling period Ts,then
all nulls at integer multiples of Tu are identically zero. If the
period is not an integer multiple of Ts, however,thefirstnull
(m<>0) actually exists between two values of m. A coarse
estimate of pitch is tolerable for many speech applications,
but is not acceptable for analysis and synthesis of music.
Compared to the small computational burden of computing
the AMDF, there is no economical method of accurately
interpolating between samples to find the true period. This
implies that the sampling rate must be sufficiently high to
yield the high accuracy required for musical applications. If
the signal is quasi-periodic (amplitudemodulated,corrupted
by noise, etc.), the nulls will never be zero, even if Tu is an
integer multiple of Ts. Theproblemofinterpolation between
lag samples to obtain an accurate pitch estimate is even
further complicated in the case of a frequency modulated
signal.
6. IMPLEMENTATION AND RESULTS
6.1 MATLAB
Here, in MATLAB, we have implemented the whole system,
with various objectives resolved step by step. The first step
taken here was the node detection along with the tuning of
the instrument. The outputs are shown under the topic of
results. Under the note detection the algorithmofAMDF was
implemented. Also the pitches of various notes were plotted
using MATLAB Function. Result using the autocorrelation
function was also successfully displayed in MATLAB.
Secondly, Raag detection was also done and errors were
found out if any. The following objectiveswerecarriedout in
MATLAB:
Objective 1: Note detection and the tuning of the instrument
Fig -6: Note and Pitch Detection
Objective 2: Error detection with respect to notes as well as
the Raag specified
Fig -7-(i): Raag Specification and Error Detection
Fig -7-(ii): Raag Specification and Error Detection

6.2 HARDWARE
The Hardware implementation of the system could be seen
below from Fig -8.
Fig -8: Hardware Setup
The results obtained on hardware are tabulated below.
Various notes were played and their frequencies were
obtained as shown in Table-2 below.
Sr.
No.
Note Played (ideal
frequencies)
Frequency Evaluated
on Hardware
1 E (329 Hz) 550 Hz
2 B (246 Hz) 533 Hz
3 G (196 Hz) 528 Hz
4 D (146 Hz) 513 Hz
5 A (110 Hz) 502 Hz
Table -2: Result Table
Here, it could be seen that the accuracy could not be
achieved. There could be various reasons for this such as
discrepancies in the algorithms, external noise in the
environment. We are working towards minimizing the
errors and increasing the accuracy.
7. CONCLUSION
The methodology used in this design is helpful and
successful. Creating the algorithm in a MATLAB .m file was
instrumental in verifyingthe outputsoverthehardware.The
Note and Pitch Detection (Frequency Extraction), Raag
Identification along with error correction was successfully
implemented in MATLAB.
The system was implemented in real time on hardwarewith
the DSK6713 kit and the results were obtained. The note
detection part was successfully implemented on hardware
but the Raag identification still remains to be implemented
from the hardware point of view.
8. FUTURE SCOPE
The future applications and extensions of this work are
numerous and varied in nature.
To extend the work of this design, advanced techniques
related to the full audio spectrum of the input audio sample
could be created. It would be interesting to note and modify
the entire spectrum of the audio sample (i.e. make a violin
sound like a trumpet) rather than focus on one single
frequency.
The accuracy of the proposed system could be improved by
making the algorithm more versatileand robustforreal time
applications. Also the DSP kit used is fairly old technology
(15 years!!!) and the support for this by Texas,Windowsand
MATLAB has gradually ceased, hence some latest kit having
support and compatibility with latest software could be
used.
REFERENCES
[1] Lyudmila Sukhostat, Yadigar Imamverdiyev, “A
comparative analysis of pitch detection methods under
the influence of different noise conditions,” Journal of
Voice, Elsevier, Volume 29, Issue 4, July 2015, Pages
410-417.
[2] Sanjivani S. Bhabad et al, “Pitch detection in time,
frequency and cepstral domain for articulatory
handicapped people,” Innovation and Technology in
Education (MITE), 2013 IEEE International Conference
in MOOC, 20-22 Dec. 2013, Jaipur, India.
[3] Rulph Chassaing and Donald Reay, Digital Signal
Processing and Applications with TMS320C6713 and
TMS320C6416 DSK, Second Edition, John Wiley and
Sons, Inc. Publication.
[4] Deborah Keil, Real-Time Data Exchange WHITE PAPER,
Texas Instruments, February 1998.
[5] Savitha S Upadhya, “Pitch detection in time and
frequency domain,”2012 International Conference on
Communication, Information & Computing Technology
(ICCICT), Oct. 19-20, Mumbai, India, July 2012.
[6] TMS320C6713 Datasheet.
https://datasheet.octopart.com/TMS320F206PZ-Texas-
Instruments-datasheet-27188.pdf

IRJET-Virtual Music Guide for Beginners using MATLAB and DSP Kit

More Related Content

What's hot (19)

Similar to IRJET-Virtual Music Guide for Beginners using MATLAB and DSP Kit (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET-Virtual Music Guide for Beginners using MATLAB and DSP Kit