 |
 |
Hi Adinda, Last night, or rather this morning, was a little hard on me and I've been making snert for the last few hours so I'm just now reading your emails. I can't access the article you reference but I'll have access at my work to all IEEE articles. I guess a big question I have is how clear you want the meaning of the lights to be. Can they be vague and is the participant/user supposed to attach meaning himself or is it made clear in advance how the lights react to the user's physiological state so the lights become like a mirror, a tool for reflection? This has consequences for the requirements for the robustness and reliability of the sensors. If the lights are supposed to change according to classified emotions the arduino will need to have a simple classification scheme. We might be able to train an LDA classifier and hardcode the resulting function in the C code. This will only work with a small number of categories (I'm thinking 4 to 8). Ah, just found the article in your other mail. Let's have a look... A system like this is similar to what I described in the previous paragraph. We are dealing with 2 instead of 4 signals and have a lot less computing power available. Also it will be very very very hard to build a goot dataset to train and test on. The last couple of months here in Barcelona I've been using these kinds of techniques to classify musical genre so I'm familiar with them. I'll look into the other chipset tomorrow and will let you know what I think. Vincent On Jan 6, 2009, at 7:38 PM, adinda van 't klooster wrote: > Thanks for all the information George! It does show that this kind > of work has to delve quite deeply into the medical and scientific, > to make sense of the data alone this is necessary.I read the article > and it does not surprise me in the least that slow breathing of the > zen monks is reflected in increased heart rate variance. As it is a > little similar to running, I have been doing breathing (pranayama) > yoga every morning for years and still break out in a sweat, even > though I don't move at all...In terms of your work, I see that your > main aim is to make people explore their own embodiment, through > exploring your interfaces which give a visual and sometimes aural > (bio)feedback. I think this is a good approach and avoids > classification. However, there will still be many mapping decisions > you take in order to give people something to go by. I agree that > the 'gaming' or goal based approach could be counter productive when > you want people to reach a meditative state, although something that > shows they are getting closer may be of use. Or perhaps they need to > forget they are trying to achieve anything altogether and so a more > abstract, immersive and peaceful audio visual feedback may be more > effective. Then where is the boundary between purely evoking through > audio visual media (as in film/video or non responsive > installations) as opposed to giving the viewer feedback of what the > effect is of the evoked state? And how to combine the two? And does > the way the artworks response have to change over time, to keep the > viewers attention and avoid boredom, or should is stay the same if > the bodily response is the same?In my own work I have similar > questions although in the Emotion Lights I am exploring the > relevance of the affective computing approach. Picard has made very > useful contributions to this field and has obtained high scores of > machine accuracy in classifying emotions from physiological input . > (see: 'Toward machine emotional intelligence: analysis of > affectivephysiological state' by Rosalind Picard, Jennifer Healey > and Elias Vyzas, MIT Media Laboratory <http://ieeexplore.ieee.org/Xplore/login.jsp?url=/iel5/34/20643/00954607.pdf?temp=x > >) <<We need more research like this!>>However, the sensors they > used are fairly invasive (some wet electrodes and belt sensor for > respiration) and I find this unsatisfactory in the context of an > artwork. I would like the viewer to be able to experience the work > in a less 'medical' way. If people have to get electrodes attached > first it changes the work to bring all sorts of associations and it > also becomes more performative (though I don't mind the performance > part on it's own). The trade off is less accuracy in the data and > hence a much less scientific starting point. This a problem I think, > but you could argue the other way. Much depends on the presentation > of the work: Does it claim to be scientific , i.e. visualize > particular emotional or cognitive states or does it merely provide a > pattern of some sorts which is physiologically related but it leaves > the viewer to discover how exactly it does relate? And if you try to > make it relate to anyone who interfaces with it, are the trade offs > simply too much? (i.e the chances are that if there is no individual > calibration, that the artwork won't make any sense to some people > other than imaginary sense... or is that the purpose of art, to > stimulate the imagination, rather than cure or teach?)Too many > questions... I know!Dare anyone try and answer some (or one) of them? > Yours, Adinda> Date: Mon, 5 Jan 2009 01:29:39 +1100> From: [log in to unmask] > > Subject: [NEW-MEDIA-CURATING] Making physiologically responsive > interactive art with heart rate variability data (for Adinda)> To: [log in to unmask] > > > on 4/1/09 10:41 PM, adinda van 't klooster at [log in to unmask] > wrote:> >> I am interested however, in how far you in your work > make an assessment of the>> person's emotions or mental state and > what exactly are you looking at and how>> does this change over > time? Which bit of information do you get when someone>> stays wired > up for 10 minutes, and why does it take that long to get it? What>> > are the certainties that can be obtained from the heart rate data > and which>> bits remain vague?> > Hi Adinda (and all the other > makers of physiologically responsive works out> there!)> > This is > going to be a very technical 'think aloud' reflection around> > biofeedback signal analysis and what it can mean - so curators > beware!> > --> > The present direction Iım heading in, is towards an > approach in which> participants explore their ability to influence > the work with whichever> emotions or breathing rhythms they want to > try out. I suppose its very> proscriptive of me (maybe its just a > phase I need to move through, and then> onto some alternate > approach) - but Iım looking for that A-HAı moment that> happens in > clinical biofeedback, when participants think "wow, I just> shifted > the patterning of my heart rhythms by thinking (feeling) about> > something melancholy/joyous", and then conversely "mmm, Iıve just > spent 20> minutes focused on my heart and breathing Iıd like to > have more moments in> my life when I can feel this settled and > present to my self" I know fully> that you cant really design a > persons response, but a goal like this can> help prioritise certain > design/copywriting and evaluational goals.> > So the interpretive > materials would need to communicate the varieties of> emotional/ > attentive orientation that can be used to influence autonomic> > nervous balance, and convey the time fames in which these > transformations> unfold from.> > Using spectrum analysis of Heart > Rate Variability to explore> psychophysiological interactions:> > > The slowest band of frequencies in a short-term HRV (heart rate > variability)> spectrum analysis is between 0.005 Hz and 0.05 Hz > (around one wave-cycle> every three minutes).> > In order to obtain > reliable results from an FFT analysis, you need at least> double > that period (nyquist rate) - so that's why it takes so long before> > you can obtain a reading, unless we can time travel!. If you want to > compare> relative amplitudes of the three key fq bins (VLF, LF and > HF) then you need> to wait till you have recorded the slowest fq.> > > There is a general agreement around the delineation of three key > frequency> bands within short term HRV spectrograms> > HF (between > 0.15 Hz and 0.4 Hz)> LF (between 0.05 and 0.15Hz) and> VLF (less > than 0.015 Hz)> ULF is connected to circadian rhythms, but can only > be measured over 24 hour> continuous recordings.> > Where it gets > hard is in the interpretation of the relative amplitudes of> these > bins in relation to each other, and what these ratios mean (i.e.> > HF:LF, VLF:LF etc.).> > Standard texts correlate increased HF to > increased respiratory sinus> arrhythmia (influence of breathing > movements on vagus nerve) and by> extension increased > parasympathetic arousal (rest-digest, calm, peaceful> etc.). > Increased LF is then correlated to increased Sympathetic arousal -> > and there's not much about VLF except that it may be connected with > body> temperature regulation.> > But this is where it gets tricky - > because if you slow your breathing down> to 6 cycles per minute > (like many forms of breath meditation) - then your> breathing will > cause your HRV to resonate at around 0.01 Hz - and its clear> to the > person doing this that they are feeling very restful and not > involved> in a typical sympathetic nervous system response.> > > Biofeedback researchers Paul Lehrer and Evgeny & Vaschillo have > referred to> this 0.01 Hz resonance as (not surprisingly) the HRV > 'resonant peak> frequency', and it is the result of resonances > between breathing, heart rate> and blood pressure. They have done > some excellent research into how this> correlates to certain breath- > based meditation techniques, and consider the> range of health > benefits these practices may provide, in relation to how> they may > be exercising certain HRV reflexes (please excuse the over> > simplifications here).> > Checkout Paul Lehrer's paper with Yuji > Sasaki and Yoshihiro Saito:> "Zazen and Cardiac Variability" (1999)> http://www.psychosomaticmedicine.org/cgi/content/abstract/61/6/812 > > this paper gives a good introduction and explores HRV spectral > data in> relation to zazen meditation practices.> > During HRV > resonance, energy is concentrated into a single peak frequency,> so > typically you'll have a peak around 0.01 hz - in the LF band (but > lower> if breathing is slower?), while the VLF and HF bands are very > low. In THIS> context - some biofeedback practitioners have > suggested the presence of> increased VLF as indicative of > rumination. I'm inclined to agree, based on> my own observations in > a clinical biofeedback session, but I've not yet> found any > published test results to support this claim.> > On reflection, I > think I'll probably end up designing my new works around> this > resonant phenomenon - which could simplify the interpretive> > requirements a bit too.> > The idea of creating interactions that > can facilitate physiological> conditions similar to those obtained > during breath based meditation is very> appealing to me (again, I'm > cautious not to reduce meditative practice to> physiological > states!).> > In this case, I'd be tracking increased 'resonance' > i.e. Simplicity of> spectral profile and resonant frequency vs. > complexity and distribution of> spectral profile. > > This would > suggest various 'conical/funnel-shaped' interaction spaces/axes,> > all directed towards or away from HRV resonance, and then responsive > to the> specific frequency of that resonance (I'm not talking here > about a physical> space in a gallery, but a mathematical space > describing the range of> possible positions one occupy within a set > of sensor data coordinates).> > How would one then describe to > audiences some approaches to interacting with> such a work without > sounding too prescriptive and new agey? A game like> challenge to > see if you can focus your heart rate pattern into a single fq,> > though a combination of slow breathing, and calmness could be one > approach -> but the idea of 'gaming' as its understood by the wider > general public,> seems at odds with the intensely meditative states > I'm encouraging people to> explore during their interaction with the > work.> > Maybe a visual map of the above mentioned conical /funnel > space, would be a> good start - but visitors would still need to > understand what HRV spectral> analysis is/means - and if this post > is anything to go by - you can see If> still got a long way to go!!> > > Using a map-like visual representation of the interaction space as > an> interpretive design, also opens the way for a less goal-oriented > engagement> (I can still remember being collected from the beautiful > virtual wilderness,> by attendants in BLAST THEORY'S 'Dessert Rain' > show, after all the other> players had collected their tokens - I > was too busy smelling the flowers!)> > The map delineates the > boundaries of the interaction space, and leaves the> participant > free to explore the space at will. Participants can also compare> > their experiences with each other by way of talking about the > different> journeys they made through this 'interaction space' > during their engagement> with the work.> > > Cheers> > > Dr. George > Poonkhin Khut> > > Mobile 0417 566 425 (International 61 417 566 > 425)> E-mail [log in to unmask]> URL http://www.georgekhut.com/> > 73 Edward Street> Darlington, NSW 2008 > _________________________________________________________________ > Express yourself instantly with MSN Messenger! Download today it's > FREE! > http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/ '(Vincent Akkermans w: http://www.triggerbangbang.nl e: [log in to unmask] a nl: Sieraadlaan 21, 2719VA, Zoetermeer a es: C/ Providencia 75 3-2, 08024, Barcelona t nl: +31 636 158 564 t es: +34 680 765 987 )