A Stress-Reducing Explorable Musical Object
by
Hannah Rhiannon Lienhard
Submitted to the Department of Mechanical Engineering
in partial fulfillment of the requirements for the degree of
Bachelor of Science in Mechanical Engineering
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
June 2019
@ Massachusetts Institute of Technology 2019. All rights reserved.
Signature redacted
A u th or ...................... ............
Department of Mechanical Engineering
May 10, 2019
Signature redacted
Certified by...
U
... Tod Machover
Muriel R. Cooper Professor of Music and Media
Thesis Supervisor
Signature redacted
A ccep ted by ...........................................................
MASSACHUSETTS INSTITUTE Maria 
OF YangTECHNOLOGYX..
Associate Professor of Mechanical Engineering
JUL 16 2019 Undergraduate Officer
LIBRARIES
-j
A Stress-Reducing Explorable Musical Object
by
Hannah Rhiannon Lienhard
Submitted to the Department of Mechanical Engineering
on May 10, 2019, in partial fulfillment of the
requirements for the degree of
Bachelor of Science in Mechanical Engineering
Abstract
In this thesis, an explorable musical object - which is meant to help reduce anxiety
and stress in the user - was designed and built. Using the program Pure Data, output
data from an Arduino was manipulated to create varied tones when several pressure,
bending, and soft potentiometer sensors were activated. These sensors were embedded
in a soft, spherical outer shell made of roto-cast silicone. The final object was fully
contained within this outer shell, as a Raspberry Pi computer was used to run Pure
Data so it did not have to be plugged in. The sound created was output through
a radio transmitter connected to a radio and headphones, again so the instrument
could be fully contained within its outer shell.
Thesis Supervisor: Tod Machover
Title: Muriel R. Cooper Professor of Music and Media
3

Acknowledgments
I would like to thank Nicole L'Huillier for her guidance and never-ending support
throughout this project and the rest of my time working in the lab. I would also like
to thank Tod Machover for his advice and support, and for giving his time to help
me make this project a reality.
Additional thanks to Graham Yeager, for helping me figure out casting, all the res-
idents of the 33rd, for being available to squish things upon request, and for being
enthusiastic about my project regardless of the current state, and to Ash Demir and
Lauren Schexnayder for being the most supportive witches I've ever met. Finally, I
want to thank Jesse Michel, for keeping me optimistic throughout this entire process.
5
6
Contents
List of Figures 9
1 Introduction 11
2 Background 13
2.1 Music as a Form of Therapy . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Sound Sculptures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3 O ther Projects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3 Mechanical Design 19
3.1 O verall Shape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Arduino and Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3 P ure D ata . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.4 C asting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.5 Sound D esign . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.6 Sensor Mapping and Interactivity . . . . . . . . . . . . . . . . . . . . 25
4 Results and Interactions 27
4.1 Final M odel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.2 Human Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5 Conclusions and Future Work 31
Bibliography 33
7
I,
8
Mil"MMIRWRIMIN"p"n -
List of Figures
2-1 'Composition #11' . . . . . . . . . . . . . . . . . . . 1 5
2-2 'Sound Sculpture' . . . . . . . . . . . . . . . . . . . 1 5
2-3 'Ollie' . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 6
2-4 'Music Shaper' . . . . . . . . . . . . . . . . . . . . . . 1 7
3-1 Shape Brainstorming . . . . . . . . . . . . 20
3-2 Concept Brainstorming . . . . . . . . . . . 20
3-3 Interaction Concepts . . . . . . . . . . . . . . . . . . 21
3-4 Breakdown of Pure Data functions . . . . . . . . . . 22
3-5 Objects used to remotely turn on Speakers in Pure Data . . . . . . 23
3-6 Full setup of Pure Data script . . . . . . . 23
4-1 Final Orb . . . . . . . . . . . . . . . . . . 27
4-2 Inner Structure .................. 29
9
10
Introduction
Music has been widely adopted as a form of therapy for conditions such as depression
and anxiety, and is commonly used by people for self-care in times of need. [1] Physical
exercise or movement is also often used to treat depression, as it gives people an outlet
for emotions that are otherwise hard to express.[2]
Many stress-reducing objects today involve fine motor movement, like fidget spin-
ners, stress balls, slime, or meditation balls. These products provide a subtle way
for people to let out pent up stress through movement, but don't allow for a more
immersive experience of stress reduction.
This thesis investigates the ways in which both of these concepts can be integrated
into one tool to help reduce stress. By creating an interactive object with both
physical and auditory feedback, users can experience multiple forms of stress-relief,
facilitating distraction from internal anxieties.
The topic of stress relief was chosen because within the MIT community stress
is a huge problem. Many students suffer from depression or anxiety, and up to 28%
of students will utilize MIT's mental health resources to help treat these conditions
throughout their time at MIT.[3] If the student body was able to utilize tools that
facilitated relaxation methods such as music therapy, they might be able to increase
student well-being and generally reduce stress across MIT's campus.
11
12
Background
2.1 Music as a Form of Therapy
Music therapy aims to aid development of a person's communicative or expressive
abilities through a combination of sound and musical meditation. There are two main
forms of music therapy: receptive music therapy, in which people listen to music, and
active music therapy, which allows people to make music. [41 Several studies have been
done that look at music therapy's effect on conditions such as depression, anxiety,
and seasonal affective disorder, among other things. It has proven to be an effective
form of treatment for many of these conditions.
One such study looked at the way it can be utilized in elder care facilities for
patients with Alzheimer's disease or dementia. It was found that listening to mu-
sic calmed patients and made them feel more at ease in situations that had other-
wise caused them anxiety or frustration. Music therapy has also been shown to aid
Alzheimer's patients in maintaining a personal identity, through exposure to music
they might remember from other points in their lives, as it can help with recalling
memories associated with a certain piece or style of music.[4]
Multiple studies have also been done on music therapy's effect on depression. One
study compared two groups of people already undergoing standard care for depression
- one group used music therapy in addition to other methods of treatment, and one did
not. Through EEG measurements of the people in both groups, it was shown that the
group using music therapy demonstrated significantly fewer symptoms of depression
13
after the study than the one that did not.[5] Active music therapy is thought to be
this effective because it allows patients to express emotions through both aesthetic
and physical experiences, and can invoke positive mood changes. It also encourages
increased awareness of one's self and surroundings, which can help patients reconnect
with themselves outside of their disorder.[4]
Other studies completed in nursing homes have shown the positive effects music
therapy can have on people suffering from seasonal affective disorder. It was found
that those completing music therapy scored lower on both the self-rating depression
scale and the Hamilton depression scale once treatment had ended than the control
group. [6]
2.2 Sound Sculptures
Interactive sound sculptures are another inspiration for this project. Allowing the ob-
server to take an active role in the progression of an artwork after it has been installed
is especially of interest. In the scope of this project, this aspect of collaboration is
especially important for stress relief, as it gives the user active engagement in that
they can control what music is created - which is a form of active music therapy.
One example of such a project is Eirik Brandal's 'Composition #11' (Figure 2-1),
which is a sound sculpture that incorporates proximity sensors. Observers can move
their hands or their whole body to change both the volume and the pitches created by
the sculpture - it behaves similarly to a theremin. The sound stops completely when
no one is close enough to it to produce sound, so passerby can't completely decipher
the work without also engaging with it.[7]
Another example is Ryan Edwards' 'Sound Sculpture' (Figure 2-2). This piece
consists of large flexible blocks containing speakers, which can be stacked and moved
around by observers. By changing the position of the blocks in relation to each other,
the music created by them is altered. Users can build new structures with the blocks,
and in doing so, change the sounds occurring around them completely.[81
14
Ii "ImiAgi
A T V
Figure 2-1: This piece by Eirik Brandal, 'Composition #11' is an interactive sound
sculpture, that uses proximity sensors to allow observers to play it as they would a
theremin. [7]
Figure 2-2: This piece by Ryan Edwards, 'Sound Sculpture' is an interactive installa-
tion, that allows users to move musical blocks around in relation to each other, thus
creating new sounds throughout the space they occupy.[8]
15
2.3 Other Projects
A variety of other therapeutic robots and devices have been created to combat de-
pression, anxiety, or dementia. One such robot is Ollie, an plush interactive otter that
responds to a user's touch. When someone strokes Ollie, it makes soothing sounds,
vibrates, or hugs the user's hand to provide positive, supportive feedback (Figure
2-3). It is sized to be approximately the dimensions of a human baby, so the user
feels as though they are caring for something that is alive and responsive to love and
affection. Ollie was an effective project, but because it was created as a part of a
senior design class at MIT, it was never sold as a product on the market. Similar
products do exist, such as Paro, a therapeutic robotic seal, but they are sold at prices
upwards of $6,000, making them inaccessible to most people looking for relief.[9]
Figure 2-3: Ollie is an animatronic otter that is responsive to human touch. It was
created as a therapy tool to help patients with dementia.[91
16
Several projects from Tod Machover and the Opera of the Future group at the
MIT Media Lab have also looked at how music can be made more accessible to the
public. One such project is the "Brain Opera", which was an interactive installation
and opera performance. The audience attending would first enter a room called the
Mind Forest, which contained several kinds of hyperinstruments they could interact
with and play. One of these hyperinstruments was the Rhythm Tree, which was a
giant percussive instrument consisting of large sacks with pads attached to them. Up
to 50 users could press the pads at a time, creating percussive music. Another was the
Melody Easel, which let users draw on a pressure-sensitive screen and outputs a single-
line melody based on what was drawn. Once the audience had explored the Mind
Forest, their input from the hyperinstruments was incorporated into the performance
aspect of the opera, which was also performed on similar hyperinstruments.[10]
Figure 2-4: Music Shapers are one of the Music Toys created for Tod Machover's
"Toy Symphony" They allow users to manipulate pieces of music through touch.[11]
Another project from the Opera of the Future that incorporates accessible music
17
is the Toy Symphony. Instruments called Music Toys were created for this project,
which are designed so that anyone can play them, regardless of age or musical back-
ground. These included Music Shapers (Figure 2-4), which were squeezable cloth
instruments created with conductive embroidery and capacitive sensing. Children
playing the Music Shapers could manipulate a piece's timbre, density, or structure -
all of which are high-level concepts they would otherwise be unable to experience -
just by squeezing the handheld instrument.[12]
Though some tools for the improvement of mental health exist, there is no product
currently on the market that utilizes music therapy to alleviate stress, anxiety, or
depression. The goal of this project has been to create an instrument that combined
both active and receptive music therapy to facilitate stress relief.
18
Mechanical Design
The desired instrument needed to be a flexible but durable structure, that would
respond to a user's touch and squeeze through sounds corresponding to their actions.
It had to be able to successfully emit these sounds somehow either through the outer
shell, or in another way that made it clear that they were coming from the object
itself. The instrument was also supposed to relax the user, so it had to be easy and
intuitive to operate, and could not emit sounds that would induce stress in users.
A large part of designing this object involved figuring out the unseen details of
the inner programs and structure. Though aesthetically the design is simple, a lot of
careful work went in to figuring out how to place sensors inside, what those sensors
should do, and how they should be interacted with.
3.1 Overall Shape
Several shapes were considered for this object, but a simple sphere was decided on
because it could maximize both the amount of sensors used and the different kind
of sensors used. This is because it has the largest, most uniform surface area of the
other shapes considered, making it ideal for embedding a variety of sensors. Other
shapes would have constraints on sensor type and amount because of their corners or
other irregularities.
19
Figure 3-1: Shape Brainstorming
A sphere also allows the object to be more mysterious. If someone is handed
an unmarked white sphere to explore, it will be harder for them to keep track of
everything they discover than it would be if the shape were something like a pyramid,
cube, or any more irregular shape. There are no landmarks on a sphere to remember
- aesthetically, the object is the same no matter how you look at it. This is also why
the orb was not marked in any way. Indicator marks like those in Figure 3-2 were
considered as a way to guide the user through the object, telling them where to press
to hear what sounds, but were decided against because it took the thrill of discovery
out of the experience.
0
Figure 3-2: Concept Brainstorming
When interacting with the object, users were expected to be able to treat the
object as they would a stress ball - which is what they were instructed to do during
testing. It would be completely compressible and flexible in multiple dimensions, so
20
it could be used without a fear of breaking or over-stretching it.
Figure 3-3: Interaction Concepts
3.2 Arduino and Sensors
A variety of sensors were needed for this project to create a wide spectrum of possible
sound outputs, that could change depending on how the users interacted with the orb.
The main characteristics sought out in sensors were: flexibility, which was necessary
so they couldn't be felt beneath the shape's outer shell; and a non-requirement for
capacitive touch, so the sensors could be activated through the outer material. This
led to the choice of three kinds of sensors - soft potentiometers, pressure sensors, and
bending sensors. Each of these was connected to an Arduino, where the output data
could be configured.
In the Arduino script, incoming data from each sensor was mapped to a specific
range of values unique to each sensor. This data was then sent to the serial port,
where it could be sent on to other programs. By mapping the output data this way,
you allow other programs taking in data from the serial port to split up the data by
sensor. This way, data from each sensor can be manipulated individually, and tailored
to each sensor's exact function - this will be done in the electronic music program
Pure Data.
21
3.3 Pure Data
Pure Data was chosen as a music software environment because of its ability to
communicate with Arduino through the serial port. It inputs the data being output
by the sensors, which was previously mapped to specific number ranges in Arduino,
with the 'comport' object. This data can then be split based on number range in
Pure Data, and then scaled to higher frequency ranges before it's sent to the speakers
as sound waves. This structure, which was used for each sensor, is shown below in
Figure 3-4.
devices(oen1
close(
comport on 96fy
split3a76 205
moses 190
mod 0 scale'180 205408E
osc~ 0
Figure 3-4: Breakdown of the Pure Data functions that act on each sensor. It splits
the incoming serial data from the comport object based on frequency, and re-maps it
to a range of higher frequencies that are then sent to the speakers.
22
loadbang
delay iOO
d dsp 1
Figure 3-5: This string of objects sends a 'bang' to the sound functionality in Pure
Data, dsp, so it turns on automatically when the program is opened. This is neces-
sary when running the program on a Raspberry Pi, because you can not turn it on
manually.
adba devces( open 10(
comport 10 9600
split3~15 45 split3 46 75 split3 76 MI]sli16 splispi3161 split3 176 205
moses 20 moses 68 moses 80 moses 11590
mod 0 mod 0 mod 0 mod mod 9 mod 9
scale 20 55 M2000i scale 50 75 400 500 scale 80 10550 sc0al0e 119 135 600 70e scale 140 175 600 909 scale 18 254 80
0
osc~ 0 osc- 0 osc- 0 osc- 9 osc~ 0 osc- a
dac-
Figure 3-6: This shows the full setup in Pure Data, where each of the six branches
coming from the comport object correspond to one sensor.
The full Pure Data script splits up each of the 6 sensors into a range of frequencies,
which are then sent to the speakers. To control when sound is on or off, a loop was
created with the 'moses' object, that directs the input based on size. If a number is
over a certain threshold, sound is let through to the osc object. Otherwise, it resets
the number to 0, which is then sent to the speakers, turning them off.
23
To get Pure Data to run remotely, a Raspberry Pi was used, which meant the
program had to run independently on startup, since there was no monitor or other
controls for the computer. This meant an object had to be added to the script that
turned on the speakers (DSP in Pure Data) automatically once the program was
loaded (Figure 3-5). It essentially sends a 'bang' to the speakers, telling them to turn
on when the program is opened. The delay you see is necessary because of the way
audio is configured in Raspberry Pi - without it, Pure Data tries to connect to the
speakers before they are fully turned on. Once this was done, a command was added
to the startup script on the Raspberry Pi that simply opened and ran the full Pure
Data script (Figure 3-6).
3.4 Casting
When designing the outer shell, casting seemed like the best option as it would allow
for a wider variety of material choice, as well as more flexibility in choosing shape.
Once a sphere was chosen as the final shape, casting also made a lot of sense because
the shell could then easily be manufactured as one solid, continuous piece. The sphere
was manually roto-cast inside a 8 inch hollow plastic sphere.
The sphere was cast from silicone, which was chosen for two reasons. For one, it
is a very flexible yet durable material, which lends itself to being squished by users
repeatedly. It's also very soft to the touch, which makes it pleasant to hold. Silicone
was also chosen because it typically has a very quick cure time (about 8 minutes),
making it ideal for manual roto-casting.
3.5 Sound Design
The sounds used in the orb were selected based on both the way the sensors them-
selves behaved, and the way people were expected to react to them. Generally, pitch
frequencies were chosen to be in range of 200-800Hz. This was for several reasons.
During testing, it was found that high pitched frequencies were more noticeable than
24
much lower ones when played together, and almost seemed to drown them out. Be-
cause of this, most pitches on the extreme ends of the frequency ranges were removed.
Most of the sensors had ranges spanning about 200Hz, so when activated some pro-
duced deeper droning sounds, and some produced higher, more 'cute' sounding pitches
that would make the orb sound almost like a purring animal. The combination of
these sounds would help the user experience joy while using the orb.
3.6 Sensor Mapping and Interactivity
Each sensor's chosen frequency range was tuned to create soothing sounds - generally,
sounds found to increase happiness or reduce stress - that varied based on the expected
user interactions. For example, the force sensors were all mapped to smaller ranges,
because the speed at which people apply pressure in this case is slow - so a smaller
range will make the pitch changes more noticeable and in tune with the person's
action. It if were a larger range, the pitch would change more rapidly and the resulting
sounds would be more chaotic and abrupt. The soft potentiometers, however, have a
larger range of frequencies, because they output data based on where they are pressed
rather than how hard they are pressed. This allows for a wider variety of pitches to
be produced from each potentiometer, and a more noticeable difference in sounds
between locations pressed on the sensor.
In designing this aspect of the orb, it had to be considered that many of the sensors
could easily be activated at once. Because of this, the sounds coming from each had
to be somewhat complimentary, but still different. If a user picked up the instrument
and poked it on one side and then the other, two somewhat different sounds should be
emitted. Otherwise, they might assume the sound is uniform regardless of how they
interact with it, and will be less motivated to continue playing with it. By varying
the way sensor data is mapped to pitch, the possibility of this happening is mostly
eliminated. This way, users can interact with the orb in any way, and always have a
different experience that urges them to continue using it.
25
26
W-W
Results and Interactions
The final outcome of this project largely met the goals initially set for the object.
The inner workings of the orb functioned as expected, and the outer shell was able
to be a fully sealed ball, allowing users to squeeze it however they desired.
~A/
Figure 4-1: The final prototype of the orb, presented with headphones hooked up to
a radio, which received sounds from the orb through a radio transmitter.
27
4.1 Final Model
The final model was also largely successful in terms of meeting the technical spec-
ifications set out initially. To assemble the entire orb, thought had to be given to
how it could maintain its flexible nature while containing the Raspberry Pi, Arduino,
battery, and the sensors.
Initially, the orb was going to have a rigid inner structure made out of acrylic, but
after initial testing it was decided against, because users could feel it underneath the
silicone. From here softer structures were considered, mainly using thin flexible foam,
however, there still needed to be some rigidity in the structure, because the force
sensors had to press against a solid object in order to work correctly. After testing
with the completed inner circuitry, it was discovered that the space occupied by the
electronics was more than enough rigid structure to allow the sensors to function
correctly, so the design was finalized with the thin foam structure.
Sound output for the device also went through several iterations before being
finalized. Initially, speakers were embedded inside the orb, so sound came directly
from the object itself. This version was very effective, because it made it seem more
like the object was a living being creating sounds in response to touch. Speakers were
moved away from largely because the orb was going to be presented in a busy space
with a lot of things going on around it, but for future iterations they will likely be
brought back. For the version presented, a radio transmitter was plugged into the
headphone jack on the Raspberry Pi, which sent audio to a radio outside the orb. This
way, the orb could be fully sealed, and the user could listen through noise-cancelling
headphones.
4.2 Human Interactions
Limited informal user testing was done with the instrument, but given the nature of
the testing no fully conclusive results could be drawn. Most people using the orb had
positive feedback - several people mentioned it was relaxing after using it for a short
28
period of time. Generally, more carefully organized user testing has to be completed
before concrete conclusions are drawn.
-W-SJU 1ilfl1 
Figure 4-2: The inner structure of the orb was made out of thin foam, which was
flexible enough that users could not feel it, but provided enough rigidity that the
force-based sensors had something to push against.
29
30
9..1.-  1, iae: 1.- .i>   ,.- .3-,  .- - -> ;- .-,  p-. .-, ,2. .rspg.-e;.- .-3,  i-, a:.  .;g: .-,,  i1.. -. .---  -.132.- - ,-a  s.r:-~y a o.iN  tl1- -:e,   ,- 'N.a ri.sM 4e- -'1  :.k  -=:"   -E - -.4.  s1'i '-:  .n'P  -    
Conclusions and Future Work
Though the orb met most of it's desired objectives, there are definitely several ways
in which the current prototype could be improved. For one, the inner electronics need
to be further refined in terms of durability. Given the nature of the instrument, it is
necessary for the rigid objects within to be designed to take a decent amount of force
in multiple directions. There were several issues during user testing where various
components got unplugged or turned off accidentally by people using the orb. These
could be avoided by designing the electronic components more carefully, so there is
more of an enclosure surrounding buttons or wires that are currently exposed within
the instrument.
Another area for improvement has to do with how the inside parts of the orb
are accessed in the current prototype. Right now, there is a slit in the side of the
silicone shell that can be resealed using superglue, which creates a strong enough seal
to keep the orb together while people are using it, but can be easily broken when the
instrument needs to be serviced. Though this is effective, it is not ideal. A potential
solution would be adding a charging port and an on/off switch to the outside of the
shell. This, paired with more durable inner electronics, would eliminate the need to
access the inside of the instrument, allowing the orb to be fully enclosed at all times.
Future iterations on this project may also explore the possibility of different sizes
and shapes for similar stress-reducing instruments. Objects large enough to be hugged
or lay down on by the user are especially of interest, because it would further the
aspects of stress-reduction that can be utilized by the instrument. Looking into
the way users react to the instrument on a more data-based level would also be
31
informative. For example, monitoring a user's heart rate while they interact with the
orb would be useful, because the way people react to specific sounds could be tracked,
and the orb could be further refined to promote relaxation.
32
Bibliography
[1] Robert Nauert PhD. Music Soothes Anxiety, Reduces Pain. Psych Central, 2018.
12] Kirsten Weir. The Excercise Effect. American Psychological Association, 2011.
[3] The Office of the Chancellor MIT. The healthy minds study 2015 survey results.
Technical report, Massachusetts Institute of Technology, 2015.
[4] Damino Laura, Jayez Sylvie, and Saladin Aurore. The Effects of Music Therapy
on Anxiety and Depression. Austin Publishing Group, 2015.
[5] JAfirg Fachner, Christian Gold, and Jaakko ErkkilAd'. Music Therapy Modulates
Fronto-Temporal Activity in Rest-EEG in Depressed Clients. Brain Typography,
2016.
[6] Xifang Liu, Xin Niu, Qianjin Feng, and Yaming Liu. Effects of five-element music
therapy on elderly people with seasonal affective disorder in a Chinese nursing
home. Journal of Traditional Chinese Medicine, 2013.
[7] Eirik Brandal. Composition #11. 2016.
[8] Ryan Edwards. Sound Sculpture. 2017.
[9] Kristina Johnson, Will Kalb, Luke O'Malley, David Orozco, Patricia P Pezzi,
Luke Plummer, Doug Sanchez, Andrew Stratton, Emily Yang, Rick Zang, Alix
de Monts, Anders Rasmussen, Carrie Liang, Chris Haid, Hazel Zengeni, Jose
Lara, Kate O'Connor, Mateo Pena Doll, Nilu Zhao, and Tim Wilczynski. Ollie.
Senior project, Massachusetts Institute of Technology, Mechanical Engineering
Department, September-December 2013.
[10] Scott Wilkinson. Phantom of the Brain Opera. Electronic Musician, 1997.
[11] Katherine Taylor. Duet for Composition and Software. The New York Times,
2013.
[12] Lisa Scanlon. Toy Symphony Maestro. MIT Technology Review, 2003.
33