Neuroinformatics meets music: Do you actually like that song or do you merely think you do?

As a neu­roin­for­mat­ics researcher, my vision is to bridge con­sumers wear­ing mobile EEG devices with online music rec­om­men­da­tion ser­vices. Our research group is already work­ing in that direc­tion, decod­ing brain­waves while lis­ten­ing to music in order to pre­dict “Like” rat­ings and then sub­mit the feed­back to music stream­ing services.

As an exam­ple, this past Sep­tem­ber we were asked to con­duct a music-EEG exper­i­ment for the ‘Music Free­dom’ ser­vice TV pro­mo cam­paign of TELIA, Nor­way’s mobile net­work oper­a­tor. By decod­ing the brain­waves of three famous Nor­we­gian artists, our task was to reveal their like­ness rat­ings for songs of var­i­ous music gen­res that they would lis­ten to.

TELIA had launched their nov­el ‘Music Free­dom’ ser­vice last sum­mer; they pro­vid­ed free data to cus­tomers who stream music to their mobiles using music rec­om­men­da­tion ser­vices such as Spo­ti­fy, Tidal and Beat. These music stream­ing ser­vices are now the “killer-apps” of a dig­i­tal ecosys­tem where all cus­tomers are con­nect­ed to the inter­net at any time and from any place using their mobile devices.

The pro­duc­tion took place in Nor­way and it all had to hap­pen fast. It was only two weeks before enter­ing the shoot­ing stu­dio when direc­tor Chris­t­ian Holm-Glad had con­tact­ed us say­ing, “Dim­itrios, we want the real thing; we want to film an actu­al exper­i­ment, sim­i­lar to what you describe in your paper”, as he explained the cam­paign concept.

TELIA’s adver­tis­ing agency, Nord DDB Oslo, had come up with an inno­v­a­tive idea. Their con­cept was to har­ness brain sci­ence to demon­strate the con­trast between lit­er­al­ly lik­ing a spe­cif­ic music genre ver­sus sim­ply think­ing of lik­ing it!

The col­lab­o­ra­tion was amaz­ing. Despite the fact that many of us were com­ing from dif­fer­ent sec­tors and back­grounds, we shared a com­mon view on future forms of dig­i­tal music. The music pros felt per­fect­ly famil­iar with the idea of decod­ing brain­waves in order to retrieve accu­rate rat­ings of user pref­er­ences dur­ing music lis­ten­ing. They could visu­al­ize along­side with us the prospect of feed­ing these rat­ings to music stream­ing ser­vices and of dynam­i­cal­ly cre­at­ed playlists that would match user’s “brain-taste”.

After sev­er­al exhaust­ing days of hard-work with Chris­t­ian and his team fol­lowed, I was final­ly on my way back from Oslo to Thes­sa­loni­ki in Greece. Reach­ing the air­port, my mind was gen­er­at­ing sce­nar­ios in the future with wear­able devices smooth­ly inter­fac­ing with online ser­vices and arti­fi­cial intel­li­gence agents proac­tive­ly antic­i­pat­ing and facil­i­tat­ing people’s desires and tastes.

“Hey Siri, what’s the weath­er like in Thes­sa­loni­ki?” I asked my phone’s A.I. assis­tant, as I was approach­ing the plane.

And then it struck me!

“Hey Dim­itrios, I have a new song for you today. Would you like to lis­ten to it?”

Dr. Dim­itrios A. Adamos is a senior teach­ing and research fel­low at the School of Music Stud­ies, Aris­to­tle Uni­ver­si­ty of Thes­sa­loni­ki (AUTh) and a mem­ber of the Neuroinformatics.GRoup.

The Study

A Con­sumer BCI for Auto­mat­ed Music Eval­u­a­tion With­in a Pop­u­lar On-Demand Music Stream­ing Ser­vice “Tak­ing Listener’s Brain­waves to Extremes” (IFIP Advances in Infor­ma­tion and Com­mu­ni­ca­tion Technology).

• Abstract: We inves­ti­gat­ed the pos­si­bil­i­ty of a using a machine-learn­ing scheme in con­junc­tion with com­mer­cial wear­able EEG-devices for trans­lat­ing listener’s sub­jec­tive expe­ri­ence of music into scores that can be used for the auto­mat­ed anno­ta­tion of music in pop­u­lar on-demand stream­ing ser­vices. Based on the estab­lished ‑neu­ro­sci­en­tif­i­cal­ly sound- con­cepts of brain­wave fre­quen­cy bands, acti­va­tion asym­me­try index and cross-fre­quen­cy-cou­pling (CFC), we intro­duce a Brain Com­put­er Inter­face (BCI) sys­tem that auto­mat­i­cal­ly assigns a rat­ing score to the lis­tened song. Our research oper­at­ed in two dis­tinct stages: (i) a gener­ic fea­ture engi­neer­ing stage, in which fea­tures from sig­nal-ana­lyt­ics were ranked and select­ed based on their abil­i­ty to asso­ciate music induced per­tur­ba­tions in brain­waves with listener’s appraisal of music. (ii) a per­son­al­iza­tion stage, dur­ing which the effi­cien­cy of extreme learn­ing machines (ELMs) is exploit­ed so as to trans­late the derived pat­terns into a listener’s score. Encour­ag­ing exper­i­men­tal results, from a prag­mat­ic use of the sys­tem, are presented.

To learn more:

• Info­graph­ic on Per­va­sive Neu­rotech­nol­o­gy: A Ground­break­ing Analy­sis of 10,000+ Patent Fil­ings Trans­form­ing Med­i­cine, Health, Enter­tain­ment and Business
• Want a music sys­tem that lifts you up when you’re feel­ing down? Meet this audio head­set with bio-sig­nal sensors
• Join the 2017 Sharp­Brains Vir­tu­al Sum­mit: Brain Health & Enhance­ment in the Dig­i­tal Age (Decem­ber 5–7th)