BandReader – A Mobile Applicationfor Data Acquisition

from Wearable Devicesin Affective Computing Experiments∗

Krzysztof Kutt

AGH University of Science and Technologykkutt@agh.edu.pl

Grzegorz J. Nalepa AGH University of Science and Technologygjn@agh.edu.pl

Barbara Gizycka AGH University of Science and Technologybgizycka@agh.edu.pl

Paweł Jemio{o AGH University of Science and Technologypaweljemiolo@gmail.com

Marcin Adamczyk AGH University of Science and Technology

Abstract

As technology becomes more ubiquitous and pervasive, special attention should be given to human-computerinteraction, especially to the aspect related to the emotional states of the user. However, this approach assumesvery specific mode of data collection and storage. This data is used in the affective computing experimentsfor human emotion recognition. In the paper we describe a new software solution for mobile devices thatallows for data acquisition from wristbands. The application reads physiological signals from wristbands andsupports multiple recent devices. In our work we focus on the Heart Rate (HR) and Galvanic Skin Response(GSR) readings. The recorded data is conveniently stored in CSV files, ready for further interpretation. Weprovide the evaluation of our application with several experiments. The results indicate that the BandReaderis a reliable software for data acquisition in affective computing scenarios.

I. Introduction

People nowadays are getting more andmore accustomed to technology that ispervasive and ubiquitous. Besides using

computers at work and at home, many own atleast one mobile device, be it a smartphone ora tablet. To address individual preferences of

∗The paper is supported by the AGH University Grant.

each user, mobile systems offer various meth-ods of personalization and customization ofthese devices. However, as they become moreand more miniature, they can be fit to piecesof hardware of a size of an accessory. Today’swearable technologies, such as wristbands, pro-vide a number of advanced sensors to moni-tor human activity and health condition. Fur-thermore, these devices could be used for de-

1

DRAFT

BandReader

livering data on human physiological statesfor affective computing applications. Affec-tive Computing (AfC) [1] is a new computingparadigm that has been gaining a lot of inter-est recently. Affective computing systems canuse a range of signals for the classification ofa person’s emotional state. Physiological signalmonitoring is one of the commonly consideredcases. For many AfC systems it is importantto use non-intrusive and pervasive technolo-gies. Therefore, wearable devices, includingwristbands, seem a promising solution.

Emotional aspect of human-computer inter-action is crucial for the process to have a nat-ural and easy flow [2]. A number of studieson usability1 has proven that careful systemdesign, mindful of user’s emotional states andresponses has a great impact on the overalluser’s performance with the application. Ourtool can simplify the implementation of practi-cal AfC software on mobile devices, while alsotaking the affective aspect of interaction intoconsideration.

The rest of the paper is organized as fol-lows: In Section II we stress the affective loop’srole in HCI, In Section III we describe selectedhardware. Then in Section IV we specify ourapplication. In Section V we provide an experi-mental evaluation of our solution and situateour approach in the context of other studies inSection VI. Finally, in Section VII we summa-rize the paper and give directions for futurework.

II. Affective Loop in

Human-Computer Interaction

Recent methods in HCI emphasize the use ofinformation regarding human emotion on theinterface and interaction level. The so-calledemotive interfaces aim at such an interface de-sign that addresses the user’s affective statedirectly, with use of various emotion inducingtechniques such as colors, photographs of hu-man facial expressions of emotions, and careful

1Usability refers to the characteristics of software thatdetermine the degree of ease and efficiency with whichthe application is used.

choice of words for slogans, titles, etc. How-ever, as technology is becoming more advancedand pervasive, a prescribed design begins tolag behind the possibilities brought by the po-tential modern conveniences. An importantaspect of AfC systems is their ability to gatherdata, process it and generate an answer dynam-ically, and in real time – a mechanism widelyknown as an affective loop. This basic conceptof AfC applications refers to the continuouschain of interactions between the user and thesoftware. As a response to the user’s oper-ations in the software, the system generatesnew content or modifies its characteristics, bymeans of colors, animations, or more subtlechanges in the parameters. Then, the user canreact to these new conditions, and so on. Inorder to implement this in the application, how-ever, one needs to adapt a specific approachand be aware of several issues that arise.

One difficulty in this context regards properdata acquisition. Oftentimes, AfC systems usemultiple sensors or complex platforms con-sisting of various devices. To ensure reliabledata interpretation, it is reasonable to con-sider signals from more than one modality –by which we mean both audiovisual data andother user’s behavior metrics. Another prob-lem is related to the temporal resolution ofdata handling. Collecting affective informationfrom the environment in asynchronous man-ner and processing it offline is not sufficientfor applying the affective loop [3] in the in-teraction. Synchronization and integration ofmultimodal and multichannel data is thereforea grave issue.

From the practical point of view, in orderto create AfC systems that support wristbandsas input devices, specific software solutionsneed to be provided. In our work, we are in-terested in ubiquitous solutions that integratewearable sensors with mobile technologies. Toacquire and process data from wristbands, ded-icated mobile applications have to be devel-oped. These applications need to address sev-eral of important requirements, such as oper-ating in real time, or the support for differenthardware solutions. In this paper we present

2

DRAFT

BandReader

such an application that we developed andsuccessfully evaluated in a number of AfC ex-periments. BandReader is an important step insatisfying these concerns, as it offers softwaresupport for wearable devices that provide phys-iological data, which is necessary to developthe affective loop.

III. Selected Wristbands for

Affective Computing

Analysis of current market state led us to selec-tion of bands that could be useful for presentedapplications: Empatica E4, Microsoft Band 2,Xiaomi Mi Band 2, Apple Watch 3 and ScoscheRhythm+. All of the devices listed here pro-vide at least a certain mode of HR monitoring,which forms a basic condition from pervasiveAfC development point of view. Consideringfitness tracking appliances developed by Fit-bit, Garmin, Polar, Xiaomi or Apple, contin-uous HR measurement is provided in eachcase. Some companies, such as Scosche, de-cided to focus just on rough cardiovascularactivity monitoring. Few developers, namelyEmpatica E4 and Microsoft Band 2, enhancedtheir wristbands with more complex biomet-rics tracking. Both devices include Electroder-mal Activity (EDA) or GSR sensors, which aidthe user’s feedback with more comprehensivephysiological body state overview. Wristbandsare a solution that provides a compromise be-tween convenience and capability. This is not,however, sufficient for achieving decent levelof affect recognition and enhancing interaction.

Easy and user-friendly body activity trackingcomes with numerous disadvantages regard-ing unusual use cases. Collected data is of-ten preprocessed before being presented to theuser. Therefore, the information that the useracquires is already interpreted in a specific,undocumented way. This is unfortunate for re-searchers who would prefer access to raw data.Information processing should be fully control-lable in order to apply best models for furtherinterpretation. In the following paragraphs weshow that most of currently purchasable (fromeveryday user perspective) biometrical instru-

ments are not supported by any satisfactorysoftware enabling to collect, store and mean-ingfully interpret data.

Xiaomi Mi Band 2 is a simple band aimed atfitness and sleep tracking. It is equipped with,among others, an optical HR sensor and a stepcounter. Xiaomi Mi Band 2 allows to checkone’s HR at a certain point of time and showsit on its display. In order to track it continu-ously, though, the device has to maintain theconnection with the official smartphone soft-ware, Mi Fit. The full functionality is broughtby using the band with one or both of the devel-oper’s applications: Mi Band Notify & Fitnessand Mi Band Tools.

Apple Watch 3 is a brand new technology forsmartwatches. This wristband is essentially theonly smartwatch we included in our analysis.The integrated components include accelerom-eter, barometric altimeter, and ambient light,gyro and HR sensors. With the preloaded Ap-ple watchOS operation system and Siri soft-ware, unlike Mi Band 2, Rhythm+ or E4, it ismeant to be a fully-developed smartwatch thatsupports SMS messaging, email, push mail andIM.

Scosche Rhythm+ is a HR sensor for exerciseintensity measurement. The distinctive featurefor this device, as claimed by the developer,is its unmatching accuracy provided with theapplication of Green/Yellow Optical Sensorstechnology. Another notable feature is its dual-mode processor, which allows simultaneousdata transmission between the Rhythm+, mul-tiple ANT+ displays (such as sport equipmentat fitness centers) and the smarthphone app(paired via Bluetooth SMART) as well. Thedevice, however is not equipped with any sortof display or screen, so any real-time data mon-itoring has to be performed with external soft-ware.

Empatica E4 is an advanced sensory wrist-band based on the technologies previously de-veloped in the Affective Computing divisionof MIT Media Lab. The device is equippedwith Blood Volume Pulse (BVP) and Electro-dermal Activity (EDA) sensors, as well as anaccelerometer and an infrared thermometer.

3

DRAFT

BandReader

To benefit from full capability of this device,the connection with user’s smartphone and of-ficial app via low energy Bluetooth Smart isnecessary. In order to access the detailed infor-mation, assembled in the CSV file generatedby the app, the user needs to download andinstall the Empatica Manager software.

Microsoft Band 2 (MS Band 2), was designedas a wearable device for health, fitness, andbasic everyday activity purposes. The sensorson board include barometer sensor, gyrometer,GPS, accelerometer, capacitive sensor, opticalHR monitor, UV sensor, ambient light sensor,and GSR sensor.

The software provided by the developers to-gether with each commercially released wrist-band is designed in a way to support the ev-eryday user. In consequence, each feature isdevised in a way that balances the efficiency,on the developer’s end, and user-friendliness,on the consumer’s part. While the specifichardware itself is often theoretically capableof providing reliable data, its full capacity isrestricted by the software that processes andinterprets the signals before presenting themto the user. Based on the available softwareoverview, we provide a critique of the presentalternatives and propose a set of features thatwould comprise to the application suitable forvarious developers needs.

Xiaomi Mi Band 2 comes along with the MiFit mobile app, which can be used both on An-droid and iOS. It allows for band managementand visualisation of historical data. Unfortu-nately, there is no open API or SDK availablefor developers. However, there are successfulattempts to “hack” the transmission done viaBluetooth and gather the data sent from theband by accessing proper bytes of raw signal.Many values can be accessed this way, whichhas been demonstrated in previous work doneby our team [4].

The first thing that makes Apple Watch 3unique is that it comes with its own opeartingsystem called Watch OS. It can be easily ex-tended using external applications. Apple pro-vides developers with full support of API andSoftware Development Kit (SDK). From the

perspective of the everyday user, Apple Watchprovides exhaustive statistical information likeamount of burned calories, sleep length ortraining progress. On the other hand, the de-vice is supported only by iOS and thanks tothat its accessibility is restricted.

Rhythm+ comes with no official mobile, PCor Web app, but there are more than one hun-dred unofficial apps available in Google Playor App Store that support it and allow activitytracking. The device uses Bluetooth SMARTand ANT+ protocols, so the user can connectit the application of choice, including other fit-ness watches, bike computers and fitness equip-ment. Producers of Scosche Rhythm+ providedevelopers with API support and Fitness Util-ity app for iOS to test third-party apps. Thissoftware can be used for accessing raw datafrom the band and generating real-time charts.Despite having a dual-mode processor, thisband is not equipped with any internal stor-age, nor does the developer allow for direct,raw data download.

Data gathered in the Empatica E4 internalmemory can be easily imported through USBand stored in the developer’s cloud platform(E4 Connect) using a software tool, E4 Manager.In turn, the platform, alongside with mobileapp (E4 Realtime app), enables an overview ofthe data. The user is able to inspect realtimecharts of gathered data, check status of batteryand duration of current session. After upload-ing data to the cloud platform, the processeddata and charts can be downloaded anytimefrom any device. Nonetheless, the quality ofdata provided and, hence, amount of informa-tion that can be derived from it, is erroneouslylimited and flawed.

MS Band 2 is supported by most mobile de-vices. In order to use its full functionality, itis mandatory to pair it via Bluetooth with themobile Microsoft Band app. Similarly to Em-patica, all of the data collected by MS Band2 is stored in the cloud and can be accessedusing the provided software. After connect-ing to Microsoft Account, the user is able tosee her activiy history, including burned calo-ries (based on amount of steps), track sleep

4

DRAFT

BandReader

activity and record changes in weight. UsingMicrosoft Health Web app, the user can accessa lot of statistical information, which is basedon the processed data. Raw data collected byMS Band 2 is practically unaccessible for theuser.

IV. Design and Implementation of

BandReader

i. Requirements for a new application

Review of current software status and our pre-vious preliminary research in AfC field [5, 6]led us to specification of a set of use cases andrequirements for the software to be a conve-nient asset for studies:

1. The technology should introduce modulararchitecture to be easily expanded withnew devices support and new functionali-ties.

2. The solution should give a possibility tosave data from all available sensors foreach band. It should take care of the factthat different devices have different sets ofsensors.

3. The applications should facilitate record-ing data from several different wristbandssimultaneously, as it is a common experi-mental scenario.

4. The software should allow synchroniza-tion with other data sources (e.g. proce-dure markers from PC). Incorporation ofLab Stream Layer system2 should be con-sidered, as it is a mature solution for thisissue.

5. There should be a possibility for arbitraryfile naming and tagging to provide morecontrol over data and to facilitate futureprocessing.

6. The data should be saved in convenientfile format (such as i.e. CSV).

2See: https://github.com/sccn/labstreaminglayer.

7. The data should be available for other ap-plications that run on the same device,i.e. there should be a service that feedsclients (other applications) with the cur-rent stream of data.

8. The application should output raw datagathered from wristband(s).

<<uses>>

WristBand

WristBandService

DataStreamer

WristBandFragment

<<uses>>

<<uses>>

EmpaticaBand

EmpaticaService

EmpaticaFragment

<<uses>> <<uses>>

MSBand

MSBandService

MSBandFragment

<<uses>>

<<uses>>

.csv

.csv

.csv

.csv

EmpaticaAPI MSBandAPI

<<uses>> <<uses>>

Figure 1: BandReader application architecture.

ii. Details of implementation

Considering the features of described hard-ware solutions, we decided to focus the im-plementation of the prototype on MS Band 2and Empatica E4.

Therefore, to ensure the greatest openness,availability and the possibility of further inde-pendent development, the BandReader appli-cation was created for the Android OperatingSystem. The developed application was writ-ten in Java language and consists of five mainmodules (see Figure 1):

• WristBand (31 LOC3) is an entity thatdescribes the type of the band and itspossible states (e.g. connecting, connected)and list of sensors available.

3Lines of Code.

5

DRAFT

BandReader

• DataStreamer (156 LOC) is a module forhandling serialization of data streams toCSV files. It converts the data to specifiedfile format and physically writes the filesto smartphone storage.

• WristBandService (287 LOC) is the mainmodule of BandReader. It supports ev-ery stage of communication with the bandwith the use of band definition writtendown as WristBand entity. Among the im-portant methods of the service there are:

– connect() connects with the wrist-band using the provided API or otheravailable method;

– startSubscription() switches theapplication to listening mode, i.e. itregisters callback methods that arecalled when new data arrives fromthe band. Current sensor readingsare now displayed in the applicationUI;

– startRecording() turns on therecording mode, i.e. received valuesare not only displayed, but also savedto CSV files;

– stream() is a generic method exe-cuted for each new sensor reading.It checks the current mode and ex-ecutes the proper actions, i.e. dele-gates data saving to DataSteamer inthe recording mode. This method canbe easily extended in the future tohandle more modes of operation, e.g.synchronization through Lab StreamLayer system or broadcasting of datato other Android applications.

It is important to note that WristBand-Service is implemented as real service,i.e. it receives and handles data even ifBandReader UI is closed.

• WristBandFragment (455 LOC) is respon-sible for UI panel for specific wirstband.With the use of methods provided byWristbandService, it allows user toconnect with the band, display the currentstatus or start recording. There is also

LoggerFragment responsible for display-ing the logs collected from DataStreamerand WristBandService.

• Utils (4 classes, 192 LOC in total) is a set ofadditional functionalities (not mentionedon Figure 1), responsible for, among oth-ers, tags handling and saving and read-ing application settings from the phone’smemory.

Figure 2: A screencapture of BandReader applicationduring the recording session.

The modular architecture of the applicationallows it to be easily extended in the future. Tosupport a new band, one must define threeclasses that inherit from the abovementionedabstract classes: (a) WristBand, with the defi-nition of band’s states and sensors, (b) Wrist-BandService, with the implementation of bandconnection and (c) WristBandFragment, with theUI panel for the user. The range of extensionof the application impacts the complexity ofimplementation. If it is a high-level change, e.g.adding synchronization through Lab StreamLayer, it will be associated only with the exten-sion of the appropriate method (e.g. stream()).Low-level functionalities that depend on thespecific device implementation, may requirechanges in individual wristband classes. De-pending on the communication methods used

6

DRAFT

BandReader

by the wirstbands’ developers, the addition ofa new device may involve a diversified numberof LOC that need to be written in order to beable to handle the instrument. In the case ofEmpatica E4, it is about 500 LOC, and MS Band2 requires as many as 600 LOC.

iii. Use Cases and Current Develop-ment

BandReader is an important part of mobile sen-sor platform we propose as a base for AfC usecases [7, 8]. Various scenarios for BandReaderin such a setting are currently under develop-ment in our team.

Data collection in AfC experiments A sub-ject takes part in experiment. She is exposedto some stimuli and at the same time her phys-iological reaction is monitored by the Ban-dReader. Signals are saved and synchronisedwith stimuli-related markers from experimen-tal PC. Such data is then processed during post-experimental analysis in order to find valuableobservations about emotional physiological re-actions, e.g. in order to prepare models foremotion prediction based on physiological sig-nals.

Music recommendations system Ban-dReader gathers live HR and GSR streamswhile the user is listening to music. As musictracks have assigned affective labels in valenceand arousal space [9], it is possible to traina neural network with the use of mobileTensorFlow library4. This model pairs currentuser state with tracks’ affective score and isthen used to propose music based on currentmood.

Geographical map of affect HR and GSR sig-nals from BandReader are transformed intoEmotional Index [10], a mono-dimensionalmeasure that reflects more positive (> 0) ormore negative (< 0) emotions. This informa-tion is stored along with current GPS coordi-

4See: https://www.tensorflow.org/.

nates and then presented for users in aggre-gated form as a chart of emotions for givenarea.

Emotion as a context informationAWARE5 [11] is a framework for build-ing context-aware applications. Together withmany other possible sources, like GPS positionor current battery usage, affective signals maybe used as an input. Such a framework, com-bined with the rule-based language designedfor inference on uncertain knowledge [8],gives a reliable base for development of moresophisticated AfC applications.

V. Experimental Evaluation of

BandReader

The Bandreader application was developed in or-der to support our practical research in the areaof affective computing. We used BandReaderto record data collected by Empatica E4 andMS Band2 wristbands in a research project re-garding the development of affective designpatterns in computer games [6]. Thanks to theportabilty of the developed solution, we arefurther extending it to provide the biofeedbackloop. Below we describe the most importantresults of the evaluation.

i. The Experimental Procedure

The experimental procedure initially com-prised of two phases, both performed on a per-sonal computer. In order to analyze the phys-iological reactions and the stimuli occurencescorrelation later, the synchronization betweenthe smartphone (with BandReader applicationrunning) and the computer was maintaned.During the whole study, the subject was wear-ing either Empatica E4 or MS Band2 whichwas paired with BandReader application viaBluetooth. In the first phase the subject is pre-sented with affective pictures [12], one by one,for a fixed amount of time. After each pic-ture, the task is to rate subjectively felt arousal

5See http://www.awareframework.com/.

7

DRAFT

BandReader

in response to the presented stimuli, on a 7-level scale from 1 to 7. The actual experimentis preceded by a short training session withneutral images. The second and last phaseinvolved playing a simple platform game, de-signed specifically for research purposes [6].The wristband continued to collect the sub-ject’s physiological data throughout the gameand the BandReader recorded it for each partic-ipant individually. The whole procedure lastedabout 22 minutes.

Figure 3: Empatica E4 wristband (on the left), alongsidethe eHealth platform and the computer used inone of our studies.

ii. Summary of Conducted Experi-ments

So far we conducted two experiments whenthe procedure involving use of BandReaderwas applied. The first experiment took placein late April 2017 (Eurokreator lab, Krakow,Poland), with 6 subjects in total. In this attemptthough, only the first phase was engaged. Thefull procedure, consisting of both phases, washeld in June 2017. This time, the study groupinvolved 9 persons.

In the studies conducted afterwards, specif-ically in January 2018 and March 2018, we fo-cused on testing reliablity of data from Em-patica E4 and MS Band 2 which was recordedwith BandReader. We decided to augment ourprocedure with two new devices: BITalino andeHealth (see Fig. 3 and 4). This time, Ban-

dReader was used in ”Multiband recording”mode, which allows to gather data from twodevices simultaneously. Both phases of pro-cedure described above were performed, and– what is more – an additional one has beenprepared. After playing the game, a relaxingpicture was showed to a user, with a very un-pleasant sound that followed. This step wasapplied to acquire data on the strong physio-logical response to a sudden affective stimuliof the subject. 18 persons participated in thestudy at that time.

Figure 4: Bitalino platform, together with GSR and HRelectrodes.

iii. Evaluation Results

As a result of the experiments, several obser-vations can be stated. First of all, BandReaderproved to be a reliable solution that supporteda connection with Empatica E4 and MS Band 2through provided API. It maintained the con-nection for the duration of the whole experi-ment, and generated proper data files (CSV).The main problems that occurred during theexperiments were related to the fact that the de-vices require special configuration, includingrepeated pairing attempts before the connec-tion takes place. However, these problems arebeyond the scope of BandReader, and are re-lated to the API implementation.

Currently BandReader effectively implementsthe core part of the requirements we have for-mulated (see Sect. IV). The application has

8

DRAFT

BandReader

modular architecture that can be easily ex-tended (Req. 1). It allows for saving the rawdata (Req. 8) from all available sensors (Req. 2)to standard CSV files (Req. 6). These record-ing may be done in parallel for many devicesat once (Req. 3). The application also givesuser a possibility to specify the file names byproviding identifiers of subjects as well as apossibility to add arbitrary tags to data (Req.5). It can be then concluded that BandReaderis a suitable tool for AfC applications, and ouraim – which was to develop a tool for affec-tive physiological data acquisition and storage– seems to be achieved.

VI. Related Research

To date, several other studies involving wear-able devices for affective computing applica-tions have been conducted. These regard eithergeneral overviews of sensors currently avail-able on the market [13, 14, 15, 16, 17], compara-tive validation between wearables and medical-quality platforms [18, 19, 20] or actual originalexperiments involving, for example, emotiondetection or recognition [18, 21, 22, 23, 14, 24].It is notable that some attempts of design-ing and developing custom measuring devicesworn on wrists are made [24, 25, 20].

In this article’s context, namely the dedicateddata recording software, it seems that not everyresearch group considered this issue, since usu-ally information about this matter is lacking inthe respective papers [17, 25, 19, 24, 26, 21, 15].Often either, official producer’s applicationsare used [13, 23], or custom-developed sen-sor is followed by custom data collecting soft-ware [18, 20]. A notable exception is providedby [14], who used their custom program forreal-time signal preview from Empatica E4Wristband, LG Watch Style and Zephyr Bio-harness Module. In the light of our state-of-artanalysis, we conclude that our application ad-dresses an important niche and successfullyprovides reliable tool for data recording andstorage from existing off-the-shelf wearable de-vices. What is more, as our software is an An-droid application, it finely fits into the current

trend of mobile and ubiquitous computing.

VII. Conclusions and Future

Work

The objective of this paper was to present a newprototype software application Bandreader foraffective computing experiments. In our work,we are planning the development of ubiquitouscomputing systems. As such, we are focusingon the use of mobile devices for data storageand processing and wearable devices for prov-ing practical measurmenets of biomedical sig-nals. In the paper we present an overview ofthe devices currently available on the market,which allowed us for selecting the ones mostsuitable for our work. We provided the de-scription of the design and implementation ofthe application, as well as results of practicalevaluation of its operation in our experiments.

As of future work, the application will beextended to support the third device: ScoscheRhythm+ wristband. In addition, implemen-tation of synchronization with the use of LabStream Layer is planned. This library provideshigh accuracy methods for combining the datastreams from various devices in local network.It will be primarily used to synchronize thedata from variuos recording devices (smart-phones, PCs and laptops) in conducted experi-ments (see also Req. 4). Finally, to support cur-rent (see Section iii) and future AfC use cases,broadcasting mechanism will be implemented(see also Req. 7). As a result, BandReader willbe considered a middle layer that gathers datafrom devices and provides the streams to otherapplications, like music recommendations sys-tem mentioned before.

References

[1] R. W. Picard, Affective Computing. MITPress, 1997.

[2] N. Thompson and T. McGill, “Affectivehuman-computer interaction,” in Encyclo-pedia of Information Science and Technology,

9

DRAFT

BandReader

Third Edition. IGI Global, 2015, pp. 3712–3720.

[3] D. Bersak, G. McDarby, N. Augenblick,P. McDarby, D. McDonnell, B. McDon-ald, and R. Karkun, “Intelligent biofeed-back using an immersive competitive en-vironment,” 2001, paper presented at theDesigning Ubiquitous Computing GamesWorkshop at UbiComp 2001.

[4] M. Wosik, “Extension of aware plat-form providing support for sensory wrist-bands,” AGH University of Science andTechnology, 2017, BSc Thesis.

[5] G. J. Nalepa, J. K. Argasinski, K. Kutt,P. Wegrzyn, S. Bobek, and M. Z.Lepicki, “Affective computing experi-ments in virtual reality with wearablesensors. methodological considerationsand preliminary results,” in Proceedingsof the Workshop on Affective Comput-ing and Context Awareness in AmbientIntelligence (AfCAI 2016), ser. CEUR Work-shop Proceedings, M. T. H. Ezquerro,G. J. Nalepa, and J. T. P. Mendez,Eds., vol. 1794, 2016. [Online]. Available:http://ceur-ws.org/Vol-1794/

[6] G. J. Nalepa, B. Gizycka, K. Kutt, and J. K.Argasinski, “Affective design patterns incomputer games. scrollrunner case study,”in Communication Papers of the 2017Federated Conference on Computer Scienceand Information Systems, FedCSIS 2017,2017, pp. 345–352. [Online]. Available:https://doi.org/10.15439/2017F192

[7] K. Kutt, W. Binek, P. Misiak, G. J.Nalepa, and S. Bobek, “Towards thedevelopment of sensor platform for pro-cessing physiological data from wearablesensors,” in Artificial Intelligence and SoftComputing - 17th International Conference,ICAISC 2018, Zakopane, Poland, June3-7, 2018, Proceedings, Part II, 2018,pp. 168–178. [Online]. Available: https://doi.org/10.1007/978-3-319-91262-2_16

[8] G. J. Nalepa, K. Kutt, and S. Bobek, “Mo-bile platform for affective context-awaresystems,” Future Generation ComputerSystems, 2018, in press. [Online]. Avail-able: https://doi.org/10.1016/j.future.2018.02.033

[9] J. Grekow, “Music emotion maps inarousal-valence space,” in Computer Infor-mation Systems and Industrial Management- 15th IFIP TC8 International Conference,CISIM 2016, Vilnius, Lithuania, 2016,pp. 697–706. [Online]. Available: https://doi.org/10.1007/978-3-319-45378-1_60

[10] G. Vecchiato, P. Cherubino, A. G.Maglione, M. T. H. Ezquierro, F. Mari-nozzi, F. Bini, A. Trettel, and F. Babiloni,“How to measure cerebral correlates ofemotions in marketing relevant tasks,”Cognitive Computation, vol. 6, no. 4, pp.856–871, 2014. [Online]. Available: http://dx.doi.org/10.1007/s12559-014-9304-x

[11] D. Ferreira, “Aware: A mobile contextinstrumentation middleware to collabora-tively understand human behavior,” 2013.

[12] A. Marchewka, Ł. Zurawski, K. Jednoróg,and A. Grabowska, “The Nencki AffectivePicture System (NAPS): Introduction toa novel, standardized, wide-range, high-quality, realistic picture database,” Behav-ior Research Methods, vol. 46, no. 2, pp. 596–610, 2014.

[13] F. El-Amrawy and M. I. Nounou, “Are cur-rently available wearable devices for activ-ity tracking and heart rate monitoring ac-curate, precise, and medically beneficial?”Healthcare informatics research, vol. 21, no. 4,pp. 315–320, 2015.

[14] S. Kye, J. Moon, J. Lee, I. Choi, D. Cheon,and K. Lee, “Multimodal data collectionframework for mental stress monitoring,”in Proceedings of the 2017 ACM InternationalJoint Conference on Pervasive and UbiquitousComputing and Proceedings of the 2017 ACMInternational Symposium on Wearable Com-puters. ACM, 2017, pp. 822–829.

10

DRAFT

BandReader

[15] H. Yates, B. Chamberlain, and W. H. Hsu,“A spatially explicit classification modelfor affective computing in built environ-ments,” in Affective Computing and In-telligent Interaction Workshops and Demos(ACIIW), 2017 Seventh International Confer-ence on. IEEE, 2017, pp. 100–104.

[16] S. Seneviratne, Y. Hu, T. Nguyen, G. Lan,S. Khalifa, K. Thilakarathna, M. Hassan,and A. Seneviratne, “A survey of wearabledevices and challenges,” IEEE Communica-tions Surveys & Tutorials, vol. 19, no. 4, pp.2573–2620, 2017.

[17] S. Rank and C. Lu, “Physsigtk: Enablingengagement experiments with physiologi-cal signals for game design,” in AffectiveComputing and Intelligent Interaction (ACII),2015 International Conference on. IEEE,2015, pp. 968–969.

[18] M. Ragot, N. Martin, S. Em, N. Pallamin,and J.-M. Diverrez, “Emotion recognitionusing physiological signals: Laboratoryvs. wearable sensors,” in International Con-ference on Applied Human Factors and Er-gonomics. Springer, 2017, pp. 15–22.

[19] C. McCarthy, N. Pradhan, C. Redpath,and A. Adler, “Validation of the empaticae4 wristband,” in Student Conference (ISC),2016 IEEE EMBS International. IEEE, 2016,pp. 1–4.

[20] S. Preejith, A. Alex, J. Joseph, andM. Sivaprakasam, “Design, developmentand clinical validation of a wrist-based op-tical heart rate monitor,” in Medical Mea-surements and Applications (MeMeA), 2016IEEE International Symposium on. IEEE,2016, pp. 1–6.

[21] M. Gjoreski, M. Luštrek, M. Gams, andH. Gjoreski, “Monitoring stress with awrist device using context,” Journal ofbiomedical informatics, vol. 73, pp. 159–170,2017.

[22] A. Triantafyllidis, D. Filos, R. Buys,J. Claes, V. Cornelissen, E. Kouidi,

A. Chatzitofis, D. Zarpalas, P. Daras,I. Chouvarda et al., “A computer-assistedsystem with kinect sensors and wristbandheart rate monitors for group classes ofexercise-based rehabilitation,” in PrecisionMedicine Powered by pHealth and ConnectedHealth. Springer, 2018, pp. 237–241.

[23] H. Koskimäki, H. Mönttinen, P. Siirtola,H.-L. Huttunen, R. Halonen, and J. Rön-ing, “Early detection of migraine attacksbased on wearable sensors: experiencesof data collection using empatica e4,” inProceedings of the 2017 ACM InternationalJoint Conference on Pervasive and UbiquitousComputing and Proceedings of the 2017 ACMInternational Symposium on Wearable Com-puters. ACM, 2017, pp. 506–511.

[24] Y. Tajitsu, “Piezoelectret sensor made froman electro-spun fluoropolymer and its usein a wristband for detecting heart-beatsignals,” IEEE Transactions on Dielectricsand Electrical Insulation, vol. 22, no. 3, pp.1355–1359, 2015.

[25] Z. Zhang, “Photoplethysmography-basedheart rate monitoring in physical activitiesvia joint sparse spectrum reconstruction,”IEEE transactions on biomedical engineering,vol. 62, no. 8, pp. 1902–1910, 2015.

[26] Z. Zhang, Z. Pi, and B. Liu, “Troika: Ageneral framework for heart rate moni-toring using wrist-type photoplethysmo-graphic signals during intensive physicalexercise,” IEEE Transactions on BiomedicalEngineering, vol. 62, no. 2, pp. 522–531,2015.

11

DRAFT