ResearchOnline@JCU

This is the Accepted Version of the following paper: Nguwi, Yok-Yen, Teoh, Teik-Toe, Ng, Choon Woo, Song, Insu, and Lin, Patrick (2013) Preliminary attraction studies. Proceedings of the 2012 8th International Conference on Computing and Networking Technology. 2012 8th International Conference on Computing and Networking Technology, 27-29 August 2012, Gyeongju, Korea (South), pp. 385-388.

Preliminary Attraction Studies Yok-Yen Nguwi1 , Teik-Toe Teoh1, Ng Choon Woo1, Insu Song1, Patrick Lin1, Siu-Yeung Cho2

1

2

School of Business (IT), James Cook University Singapore, 600 Upper Thomson Road, Singapore 574421

Division of Engineering, The University of Nottingham Ningbo China

Abstract— Physical beauty ensnares hearts, capture minds, and attract attention. In 1992, Naomi Wolf [1] set aside centuries of speculation when she said that beauty as an objective and universal entity does not exist. Attempts to measure beauty quantitatively have been made by investigators in psychology, arts and image analysis, and more recently in oral and maxillofacial surgery [4-5]. However, the publications related to analyzing facial attractiveness by computational intelligence are numbered. In this paper, we present an attempt to quantify facial attractiveness through the use of EEG response. This preliminary study aims at describing the data we obtained through questionnaire and EEG recording. The EEG response provides some encouraging differences which acts as a launching point towards the next step of more detailed EEG data analysis using computational intelligence. Keywords-face; attractiveness; eeg; questionaire

I.

INTRODUCTION

Physical beauty ensnares hearts, capture minds, and attract attention. In 1992, Naomi Wolf [1] set aside centuries of speculation when she said that beauty as an objective and universal entity does not exist. "Beauty is a currency system like the gold standard. Like any economy, it is determined by politics, and in the modern age in the West it is the last, best belief system that keeps male dominance intact." It has long been believed that the concept of facial beauty is variable and subjective to race, culture or era. However, psychological and medical sciences state that there is a timeless, aesthetic ideal facial beauty based on facial proportions. Scientific studies also reveal that the concepts of a ‘‘beautiful face’’ are not learned but seemto be ‘‘hard-wired’’ into our mind from birth [2]. Moreover, cross-cultural investigations on facial beauty show that different groups have similar perceptions of facial beauty, and a universal concept of beauty may be defined throughout different races, cultures and eras [3]. Attempts to measure beauty quantitatively have been made by investigators in psychology, arts and image analysis, and

more recently in oral and maxillo-facial surgery [4-5]. The most famous of these are based on the Golden Proportions (derived from the Golden Ratio, or phi: 1.61803) [6] and the Facial Thirds [7]. However, measuring facial beauty remains a challenging task. The existing approaches either lack general confirmation from a significant pool of human referees, or require several cumbersome manual measurements, or both. Instead, it is important that the approach be based on experiments with sufficient human referees and automated image analysis tools. Such an automated and objective beauty classifier could be extremely useful in several applications such as plastic surgery (for predictive evaluation of facial beauty before surgical procedures), thecosmetic and entertainment industries, and virtual media. There has been some attempt to explore male facial beauty [8]. In his paper, Peseo describes the similarities and the slight differences of ratios and measurements for either gender to be considered attractive. He similarly bases his analysis on the Golden Proportions and Facial Thirds rules and adds several ratios and criteria to them derived from other canons. Eventually, extending the analysis and classification to make facial beauty by a similar automated analysis of proportions seems feasible [9]. David et. al. [10] maps faces onto a human face shape space, and then quantitatively analyses the effect of facial geometric features on human facial beauty by using a similarity transformation invariant shape distance measurement and advanced automatic image processing techniques. The experiment reveals that human face shapes lie in a very compact region of the geometric feature space and that female and male average face shapes are very similar. They further demonstrate that a face can become more beautiful by making its geometric feature getting obviously closer to the average face shape, but if its distance to the average face shape is already relatively small, deforming it further toward the

average face attractiveness.

shape

cannot

effectively

improve

its

In this paper, we present an attempt to quantify facial attractiveness through the use of EEG response. This preliminary study aims at describing the data we obtained thorough questionnaire and EEG recording. The EEG response provides some encouraging difference which acts as a launching point towards the next step of more detailed EEG data analysis. Section II describes some background information on the importance of EEG in similar field. Section III gives information about the experimental and data. II.

BACKGROUND

Electroencephalogram (EEG)-based facial response is a relatively new field in the affective computing area with challenging issues. The EEG recording is made from electrodes attached to the surface of scalp measuring the differences from reference electrode. Action potentials in axons contribute little to scalp surface records as they are asynchronous and the axons run in many different directions [16]. Surface records are the net effect of local postsynaptic potentials of cortical cells. These may both excitatory and inhibitory. Each pyramidal cell acts as a radially orientated dipole. The total contribution at the surface depends on the staggered summation in space and the degree of activity at any given time. The orientation of the dipole reverses when most of the input at the dendrites is inhibitory. The first recording of the electric field of the human brain was made by the German psychiatrist Hans Berger in 1924 in Jena [17]. He gave this recording the name electroencephalogram (EEG). Guillaume et. al. [11] reason the use of EEG for emotion assessment which utilizes spontaneous and less controllable reactions as provided by physiological signals. Physiological signals can be divided into two categories: those originating from the peripheral nervous system (e.g. heart rate, ElectroMyogram - EMG, galvanic skin resistance-GSR), and those coming from the central nervous system (e.g. ElectroEncephalograms-EEG). In recent years interesting results have been obtained with the first category of signals ([12], [13]). Very few studies however have used the second category [14], even though the cognitive theory states that the brain is heavily involved in emotions [15]. They presented a study that makes use of classification techniques on features extracted from physiological signals to assess the arousal dimension of emotions. The inverse problem in electroencephalography [16] aims at reconstructing the underlying current distribution in the human brain using potential differences and/or magnetic fluxes measured directly, or at a close distance, from the head surface. Source localization can be separated into two main problems – forward and inverse modeling. Forward modeling serves for computing surface potential distribution from known current distribution in the human brain. For linear inverse models, forward problem is used for computing so called lead field matrix – the forward solution at every grid point of usually homogeneously discretized brain volume. For

parametric modelling methods, as the estimated dipoles move through the human brain, forward solution is repeatedly recomputed to estimate scalp distribution [18]. Further elaboration can be found at [19-21]. For the biological signals of interest in EEG, the timederivatives of the associated electric and magnetic fields are sufficiently small that they can be ignored in Maxwell’s equations. The static magnetic field equations are ∇ × b(r) = μ0j(r) and ∇ ⋅ b(r) = 0 , i.e. the curl of the magnetic field at location r is proportional to the current density j(r) , and the divergence of the magnetic field is zero. We are interested in the current density in a closed volume of finite conductivities. Outside this volume the conductivity and current density are zero. The integral equation relating b(r) and j(r) is the integral form of the Biot–Savart law 𝑏 𝑟 =  

!! !! !

𝑗 𝑟! ×

! !!

𝑑𝑟 !

(1)

where d = r – r’ (with magnitude d) is the distance between the observation point r and the source point r', and the integration is carried out over a closed volume G. Inverse solution does exactly the reverse process. Starting from known scalp distribution of potential, it estimates the underlying current distribution within the brain volume. Unfortunately, in contrast to forward problem, it has not unique solution – for the same surface potential distribution, there are infinite combinations of current sources within the human brain. III. THE EXPERIMENTS AND DATA This section details the creation of a database in relation to responses towards face attractiveness. There are two sets of data being collected during the experiment. The first set of data is in the form of questionnaire, whereas the second set of data is collected via electroencephalogram (EEG) device. A set of slide shows encompassing 50 subjects’ photos are flashed to each participant. Each participant is first presented with the slide shows that last about 8 minutes. The relevant EEG points are connected to collect signals from different part of brain. The second part of the experiment requires participants to answer some demographics information about him/her and how they define attractiveness. The questionnaire also requires the participants to rate the faces of the same 50 subjects’ images and list down the top 3 facial features that are appealing to them and rate it accordingly. Each of these images has been extensively evaluated by participants, providing valence/arousal values as well as ensemble means and variances. However, as observed during experiments, feelings induced by an image on a particular participant can be very different from the ones expected. This is likely due to difference in past experience. The EEG is a measure of brain waves. The aim of including EEG in this study is to study the activity in the brain when the participants are viewing faces of different subjects. The EEG system has 8 channels that connect to left and right

hemisphere. The EEG system is running at 256Hz, so there are 256x8 channels per seconds. For 5 seconds of viewing a photo, there are 5x256x8 data per second that makes it too large for the purpose of analysis. For this reason, we only extract the data in the center interval that would give sufficient information about the response for a subject. For each experimental recording, the participant equipped with the EEG sensors was sitting in front of a computer screen in a bare room relatively immune to electromagnetic noise. A dark screen was first displayed for 3 seconds to “rest and prepare” the participant for the next image. An image was subsequently displayed for 5 seconds. This process is continued for all the 50 images as shown in Figure 1. Half of the subjects are images of models or celebrity obtained from search engine. The other halves are selected from CMU [22] emotion database’s neutral faces. Finally, the participant was asked to assess the response of his/her value on attractiveness. The questionnaire was not limited in time to allow for a resting period between images. There are a total of 60 participants participated in the experiment. They come from various countries like America, Taiwan, China, New Zealand, Australia, Indian, Indonesia, Myanmar, Vietnam, and Singapore. All of them are enrolled students in the university studying different disciplines like Psychology, Information Technology, and Business. They age from 18 to 37 years old. The first set of questionnaire data shows that only 17% of participants find that the most important factor in attractiveness come from facial attractiveness. The main source of attractiveness comes from personality, about 31%. The third important factor is intelligence, about 11% of participants value intelligence as the main factor of attractiveness. Other minority factors include body attraction 9%, interpersonal skills 7%, social economic status 5% and other factors 20%. The other question which compares behavioral and physical attractiveness reveals that only 3 out of 60 person values physical attractiveness more than behavioral. The rest of them are mainly attracted by behavioral attractiveness. Figure 3 shows the snapshots of 8 channels EEG response for 1 sec of recording. Figure 3(a) shows response of participant 11 towards the attractiveness of subject 13. Figure 3(b) shows response of participant 12 towards the same subject. The green colour channel 4 is shown to have significant differences. Further analysis on EEG response on attractiveness will be published in future publications. IV.

same subject. Further computationally analysis will be the future works that embarks from this work. REFERENCES [1] [2] [3]

[4] [5]

[6] [7] [8]

[9]

[10]

[11]

[12]

[13]

[14]

[15]

[16]

[17]

CONCLUSION

A set of database measures the response of facial attractiveness is presented in this paper. There are two sets of data being collected during the experiment. The first set of data is in the form of questionnaire, whereas the second set of data is collected via electroencephalogram (EEG) device. The aim of including EEG in this study is to study the activity in the brain when the participants are viewing faces of different subjects. The questionnaire data shows that only 17% of participants find that the most important factor in attractiveness come from facial attractiveness. The main source of attractiveness comes from personality, about 31%. The EEG response also reveals that different participants are viewing the

[18] [19]

[20]

[21]

[22]

Naomi Wolf, The Beauty Myth: How Images of Beauty Are Used Against Women (New York: Anchor, 1992), 291 Larrabee, W., 1997. Facial beauty: myth or reality? Archives of Otolaryngology-Head and Neck Surgery 123, 571–572. BBC Science—the Human Face, http://www.bbc.co.uk/science/humanbody/humanface/beauty_golden_m ean.shtml (August, 2002). Bell, A., 1997. The Definition of Beauty, Nature, October/November Issue. Cunningham, M.R., Roberts, A.R., Barbee, A.P., Druen, P.B., et al.,1995. Their ideas of beauty are, on the whole, the same as ours. Journal of Personality and Social Psychology 68, 261–279. Huntley, H.E., 1970. The Divine Proportion: A Study in Mathematical Beauty. Dover Publications, New York. Farkas, L.G., 1994. Anthropometry of the Head and Face, second ed. Raven Press, New York. Peseo, G., 2003. The ‘‘Beauty’’ of homo sapiens: standard canons, ethnical, geometrical and morphological facial biotypes (part three). Virtual Journal of Orthodonics 5 (2) ISSN–1128-6547. Hatice Gues and Massimo Piccardi, Assessing facial beauty through proportion analysis by image processing and supervised learning, Int. J. Human Computer Studies 64 (2006), pp. 1184-1199. David Zhang, Qijun Zhao and Fangmei Chen, Quantitative analysis of human facial beauty using geometric features, Pattern Recognition 44 (2011) pp. 940-950. Guillaume Chanel, Julien Kronegg, Didier Grandjean and Thierry Pun, Emotion Assessment: Arousal Evaluation Using EEG's and Peripheral Physiological Signals. July 17 th – August 11 th , Dubrovnik, Croatia ⎯ Final Project Report Proc. Int. Workshop on Multimedia Content Representation, Classification and Security. (2006) pp 530-537 C.L. Lisetti and F. Nasoz, "Using Noninvasive Wearable Computers to Recognize Human Emotions from Physiological Signals", Journal on applied Signal Processing, Hindawi Publishing Corporation, 2004, pp. 1672-1687. B. Herbelin, P. Benzaki, F. Riquier, O. Renault, and D. Thalmann, "Using physiological measures for emotional assessment: a computeraided tool for cognitive and behavioural therapy", in 5th Int. Conf on Disability, 2004, Oxford. V. Bostanov, "Event-Related Brain Potentials in Emotion Perception Research, Individual Cognitive Assessment, And Brain-Computer Interfaces", PhD Thesis, 2003. D. Sander, D. Grandjean, and K.R. Scherer, "A systems approach to appraisal mechanisms in emotion", Neural Networks, Elsevier, 2005, pp. 317-352. Josef Rieger, Lenka Lhotská, Vladimir Krajca, Milos Matousek: Application of Quantitative Methods of Signal Processing to Automatic Classification of Long-Term EEG Records. ISBMDA 2004: 333-343 Berger's invention has been described "as one of the most surprising, remarkable, and momentous developments in the history of clinical neurology". David Millet (2002), "The Origins of EEG" International Society for the History of the Neurosciences (ISHN) EEG Signal Analysis, Josef Rieger. Doctoral Thesis 2006 pp. 12-14. MOSHER, J. C., LEAHY, R. M., LEWIS, P. S. EEG and MEG: Forward Solutions for Inverse Methods. IEEE Transactions on Biomedical Engineering, March 1999, vol. 46, no. 3, p. 245-259. MOSHER, J. C., LEAHY, R. M., LEWIS, P. S. Matrix Kernels for the Forward Problem in EEG and MEG. Technical Report LA-UR-97-3812. Los Alamos, 1997. SARVAS, J. Basic mathematical and electromagnetic concepts of the biomagnetic inverse problem. Physics in Medicine and Biology, January 1987, vol. 32, no. 1, p. 11-22. The_face_research_group. CMU Image Data Base [Electronic Version],from http://vasc.ri.cmu.edu/idb/html/face/

Figure 1 The 50 subjects of attractiveness study

(a) Response of 8 Channels EEG for 1 sec of image(13) for participant 11

Figure 2 Important Factors in Attractiveness

(b) Response of 8 Channels EEG for 1 sec of image(13) for participant 12 Figure 3 EEG response on subject 13