about breathing_time

For the TIK festival documentation I wrote an article about breathing_time:

Background and concept

Breathing_time was conceived as part of the Time Inventors Kabinet[1] project for which I was an invited artist. The idea behind this project was to use different ecological input for creating new notions of time. Right from the start I had the idea to work with physiological data as input for a new time. Can we make time more personal if it is driven by our own body? Can we change our perceptions of time through growing awareness of the way our body functions? These were thoughts that motivated the work.

The concept of the windclock[2] was a central theme in the TIK project the most obvious physiological data to work with was breathing.

Early on in the project I had the idea of representing this personal data in a direct way using analogue techniques like drawing. I experimented a lot with ink and stains and made a hand driven drawing machine that drew a line of various thicknesses depending on the speed of breathing. I drew inspiration from Japanese calligraphy techniques, especially ensō[3]. While the idea of ink stayed it changed from analogue to digital: an animation with sound to represents the breath flow.

I wanted to work with a group of five people breathing at the same time and explore if becoming aware of someone else’s breathing pattern would influence your own and if we could reach a certain entrainment, our own rhythm. This resulted in two performances performed at the TIK festival.


I build a custom device, the breathCatcher, using the JeeLabs RBBB Arduino[4] and the Modern Device Windsensor[5] and USB Bub[6]. The device is cone shaped to capture the breath flow in both directions. The wind sensor is placed in the opening of the cone. The cone should be worn over the nose and mouth. Breathing in and out through the nose is required. A felt ring protects the face from the sharp paper edge. A felt container at the bottom holds and protects the microcontroller. The paper device is connected to a PC by a cable using a USB-to-serial connection.

Sensor platform

For working with the sensor data I used the CommonSense platform[7]. I was sponsored by the Sense-os, the creators of that platform. CommonSense is an advanced online platform with comprehensive API[8] for working with sensor data. After creating an account you can create sensors, five in my case, and upload to and download data from the server. Different queries are possible and basic visualisation is available. That comes in very handy when you are developing.

I received a lot of help from Sense-os with connecting to the API and querying the database. All data is exchanged in JSON format which is very particular about quotes, which made it hard to work with.

For them the challenge lay in the near real time service of sending and receiving five times ten data points per second. I was advised to use a cable instead of Wifi to ensure minimal data loss.


I wrote custom software, drawingBreath, in Processing[9]. I used some native Java and a few extra libraries and classes.[10] This software performs all the connections with the CommonSense API. It uses several timers to keep the tasks of sending and receiving data separated.

For 60 seconds the software calibrates all five devices so as to be able to detect the direction of the breath flow. Using the temperature sensor was very useful for that purpose.

After the breath flow has been calibrated the animation starts. Each of the five participants is represented by a ‘brush tip’ which will start to draw a circle. Going counter clockwise in red represents breathing in, the blue dot moving clockwise represents breathing out. The radius of the circle is determined by the strength of the breath flow as is the size of the tip and its’ colour intensity. In between breaths the drawing clears to start again.

Other software used in, and in aid of this project was Csound, Skype, Dropbox (view below) and NTP[11]. The latter was very important as the timestamp for every breath data point should be the same.

Adding sound

My friend Richard van Bemmelen, a musician and programmer kindly offered to help me add sound to the animation. My idea was to create a bamboo wind chime with our breaths. Creating a sound only when the breath status changed from in to out or vice versa. Richard is an advanced user of Csound[12] and wanted to use that program. As bamboo already exists as an Opcode[13] we could quickly start. The sound produced by Csound wasn’t the rattle of sticks but a far more beautiful flute-like sight. The pitch depends on the value of the breath flow data. To make everything work on all the participants’ PCs Csound had to be installed. A custom .csd file which defines the settings for the synthesizer is placed in that folder. To make starting of the sound part easy Richard created a batch file that would start Csound and make it wait for messages from Processing. For communicating with Csound the oscP5 library[14] was used in Processing. A message with the breath value was send whenever the breath status changed.

The performances

breathing_time was a networked performance. I’ve selected five people from different nationalities to partake in the experiment. With that I wanted to underline the universal character of breathing. From five different locations these five people would create sound and visuals using only their breath. Because of the drawingBreath software all participants saw the same animation and heard the same sounds. This output could act as feedback for them. I was in Brussels performing for an audience that saw and heard the same things as the participants.

One thing that took a lot more effort then anticipated was preparing the participants for the actual performances. To test the server and different versions of the software we had planned four test sessions at the start. But first all software had to be installed on the different computers. Right at the beginning I had to move everybody to the Windows platform as running the Processing application made on a Windows PC on a Mac appeared to be a hassle. Also the drivers for the USB Bub were absent for the Mac.

Having equipped two participants with my old laptops we could start testing. The Sense-os server did a very good job. The main problem was instructing everybody and making sure that the software and Csound updates were put in the right folders. I used Dropbox[15] to supply updates and manuals but even that was hard for some people. Through Skype I gave live instructions and could answer questions of all participants at the same time. After a good final rehearsal it was time for the real thing.

The performances started with each participant introducing him/herself in a pre-recorded sound file in both their mother tongue and English. At exactly 19:00 hours everybody would start their drawingBreath program and calibration started as the introductions continued.

Our assignment for the performances was: relax and breath naturally. Try to detect your own breath circle and see if you can leave some time between each breath. If this moment of in between breaths would coincide the screen would be cleared and we would have reached some sort of communal breathing.

The most important thing I learned from the performances is that breathing is a very personal thing that isn’t easily manipulated. This shows very well from the CommonSense logs where you can see the breathing pattern almost as a signature.[16] Our breathing gaps didn’t coincide but the different movements of the breath flows was interesting to watch.

I also realised that although the performances went reasonably well this is just the beginning. There are so many things that could be improved for which I just lacked the time. Enthusiastic reactions have brought to me new ideas of working with the concept. I’m considering creating an online community to improve the hard- and software. To breath together online and explore the idea of creating a communal “breathing time” further.


drawingBreath software (Processing & Java), breathCatcher hardware (Arduino RBBB, Modern Device Wind sensor, USB Bub, USB cable, paper, felt, elastic band), sensor platform (CommonSense API), sound (Csound & Processing)


Concept, design, development & programming: Danielle Roberts

Sound: Richard van Bemmelen

CommonSense API: Sense-os

Participants: Adriana Osorio Castrillon, Lorenzo Brandli, Mieke van den Hende, Tomoko Baba

Location: Imal, Brussels

Also made possible by OKNO

Blog: http://www.numuseum.nl/blog/category/breathing_time/

[1] http://timeinventorskabinet.org/

[2] http://www.timeinventorskabinet.org/wiki/doku.php/windclocks

[3] en.wikipedia.org/wiki/Ensō

[4] http://jeelabs.com/products/rbbb

[5] http://shop.moderndevice.com/products/wind-sensor

[6] http://jeelabs.com/products/usb-bub

[7] http://www.sense-os.nl/commonsense

[8] http://www.sense-os.nl/api-console

[9] http://processing.org/

[10] Processing serial and net, guicomponents GTimer class, org.json and Java.net.URL and URLConnection classes

[11] http://www.meinberg.de/english/sw/index.htm

[12] http://www.csounds.com/

[13] http://www.csounds.com/manual/html/bamboo.html

[14] http://www.sojamo.de/libraries/oscP5/

[15] www.dropbox.com

[16] http://www.numuseum.nl/blog/2012/05/11/performance-11-5/

breathing sequence slow

I’ve managed to program the first representation of breathing-time. This is
a sequence of the way the drop changes when I’m actually breathing while
wearing the stretch sensor. The dept of the breath determines the diameter
and alpha value of the ‘drop’. The breath rate determines the blurring and
horizontal position of the drop. Slow but deep breathing results in a soft
and blurry spot. I’ll post an example of fast breathing later.


discovering breath


I’ve upgraded my code for the stretch sensor graph in Processing. The sensor
outputs numbers. When I breath in or out the number gets lower. Detecting
breathing activity isn’t as easy as looking for a numbers below 200 for
example. Because over the time of wearing the sensor the whole range of
numbers starts to shift going either up or down. What remains is the sharp
decrease when breathing in. So now I’m comparing two averages of 5 rounds of
serial port activity each. When their difference is over 107% I know I’m
breathing in.

stretch sensor visualisation

I’ve upgraded the first circuit with the potentiometers and the results look
promising. I used the Processing sketch that comes with the Arduino to make
a graphic of my breathing activity. I’ve tagged the different regions in the
graph so it’s easy to follow the movement of the breath. Btw the circuit is
constructed in such a way that the resistance decreases when the sensor is
stretched (breathing in.)
For this experiment I put the belt with the sensor rather tight around my
waist and I wasn’t talking. Talking makes the ‘not breathing’ part more
ragged but you can still clearly see when I’m breathing in and out.