Abstract: Recent progress in 3D technology depends greatly on computer innovation. Three-dimensional (3-D) television sets are already available in the market, and they are becoming increasingly popular among consumers. However, asthenopia and motion sickness are potential adverse effects of 3-D movies. Visually induced motion sickness (VIMS) is caused by sensory conflict, i.e., the disagreement between vergence and visual accommodation while viewing stereoscopic images. VIMS can be analyzed subjective and physiologically. The objective of this study is to develop a method for detecting VIMS. The authors quantitatively measured head acceleration and body sway during exposure to a two-dimensional (2-D) image and conventional three-dimensional (3-D) movie. The subjects wore head-mounted displays (HMDs), and they maintained the Romberg posture for the first 60 s and a wide stance (midlines of the heels 20 cm apart) for the next 60 s. Head acceleration was measured using an active tracer at a sampling frequency of 50 Hz. The Simulator Sickness Questionnaire (SSQ) was completed immediately afterward. Statistical analysis was applied to the SSQ subscores and to each index for stabilograms. Transfer function analysis indicates that the acceleration of the head in the anterior-posterior direction while watching a 3-D movie can affect lateral body sway, thereby causing VIMS.
　　Key words: Visually induced motion sickness, stabilometry, sparse density (SPD), head acceleration, transfer function analysis.
　　1. Introduction
　　The human standing posture is maintained by the body’s balance function, which is an involuntary physiological adjustment mechanism termed the righting reflex [1]. To maintain a standing posture when locomotion is absent, the righting reflex, centered in the nucleus ruber, is essential. Sensory signals such as visual inputs and auditory and vestibular inputs as well as proprioceptive inputs from the skin, muscles, and joints are involved in the body’s balance function [2].
　　Stabilometry has been used to evaluate this equilibrium function qualitatively and quantitatively. A projection of a subject’s center of gravity onto a detection stand is measured as an average of the center of pressure (COP) of both feet. The COP is traced for each time step, and the time series of the projections is traced on an x-y plane. By connecting the temporally vicinal points, a stabilogram is created. Several parameters such as the total locus length (LNG), area of sway (ARS), and locus length per unit area(LNG/ARS) have been proposed to quantify the instability involved in the standing posture, and such parameters are widely used in clinical studies [3]. Firstly, the LNG shows the overall length of the COP locus in the measurement time as
　　x y?
　　corresponds to time series of the COP. Secondly, the area of sway (ARS) shows the inner area surrounded by the outline of the greatest oscillation of locus. A stabilogram was divided equally by 120 units toward the circumference (360°). We herein define the outline of the greatest oscillation of locus in i-th unit as
　　iiirxy??. Lastly, the parameter (L/A), in particular, depends on the fine variations involved in posture control [1]. This index is then regarded as a gauge for evaluating the function of the proprioceptive control of standing in human beings. However, it is difficult to identify the decline in equilibrium function by utilizing the above-mentioned indices and measuring patterns in the stabilogram. Large interindividual differences might make it difficult to understand the results of such a comparison.
　　The analysis of stabilograms is useful not only for preventing elderly people from falling [4] but also for achieving control of upright standing by two-legged robots. Recent studies suggest that maintaining postural stability is a major goal of animals [5] and that they experience sickness symptoms in circumstances where they have not acquired strategies for maintaining their balance [6]. Riccio and Stoffregen argued that motion sickness is not caused by sensory conflict but by postural instability, although the most widely known theory of motion sickness is based on the concept of sensory conflict [6-8]. Stoffregen and Smart(1999) reported that the onset of motion sickness may be preceded by significant increases in postural sway[9].
　　The equilibrium function in humans deteriorates when viewing three-dimensional (3D) movies [10]. This visually induced motion sickness (VIMS) has been considered to be caused by a disagreement between vergence and visual accommodation while viewing 3D images [11]. VIMS can be measured by subjective and physiological methods, and the simulator sickness questionnaire (SSQ) is a well-known subjective method for measuring the extent of motion sickness [12]. The SSQ is used herein for verifying the occurrence of VIMS. The following parameters of autonomic nervous activity are appropriate for the physiological method: heart rate variability, blood pressure, electrogastrography, and galvanic skin reaction [13-15]. A wide stance (with midlines of the heels 17-30 cm apart) reportedly results in a significant increase in the total locus length in the stabilograms for individuals with high SSQ scores, while the length in those of individuals with low scores is less affected by such a stance [16].
　　By using the SSQ and stabilometry, in this study, we examined whether the VIMS was induced by a stereoscopic movie. We then investigated effect of the stereoscopic movie on the body sway and head acceleration. Based on analysis of these time series, the objective of this study is to develop a method for detecting VIMS. In this paper, relationship between them is also discussed by using transfer function analysis. In order to stabilize visual field in human beings, it is necessary to control head posture. The head and body posture is controlled by labyrinthine reflex[17] and optical righting reflex, respectively (Fig. 1). Some of the reflexes and neuroanatomy have been defined and illustrated separately [18]. However, collective reflex and their interactions have not been elucidated although a corporative effect was seen in
　　 their relationship between head movement and the movement of the center of gravity [19]. By showing a stereoscopic movie to the subjects, Takeda et al. verified that there is a corporative correlation between the head movement and the sway [20]. We herein assume that the input signal, x(t), is the head acceleration in the transfer system to control the body sway as shown in Fig. 2.
　　In this study, we examine the hypothesis; the motion sickness induced by 3-D images affected the body sway and head acceleration, and these control system changes. Especially, we took attention to form of stabilogram. The paper is organized as follows: Section 2 describes materials and methods that are used in this experiment. In section 3, we analyzed head accelerations and stabilograms of the resting state. They
　　were statistically compared with those observed when subjects viewed the 3-D movie. Section 4 discuses the comparison between the head accelerations and the body sway in accordance with transfer function analysis. Finally, section 5 concluded that VIMS of subjects that have a tolerance to motion sickness could be detected especially by calculating our SPD as an index of stabilograms.
　　2. Materials and Methods
　　2.1 Participants
　　Thirteen healthy subjects (age, 23 ± 6.2 years) voluntarily participated in this study. All of them were Japanese and lived in Nagoya and its surrounding areas. They provided informed consent prior to participation. The following subjects were excluded from the study: subjects working in the night shift, those dependent on alcohol, those who consumed alcohol and caffeine-containing beverages after waking up and less than 2 h after meals, those who had been using prescribed drugs, and those who may have had any otorhinolaryngologic or neurological diseases in the past (except for conductive hearing impairment, which is commonly found in the elderly). In addition, the subjects must have experienced motion sickness at some time during their lives.
　　2.2 Materials
　　Images were presented on a HMD (iWear AV920; Vuzix Co. Ltd.). Using a stabilometer (G5500; Anima Co. Ltd.) and active tracer (AC-301A; GMS Co. Ltd.), we measured body sway and head acceleration, respectively. 2.3 Design
　　We ensured that the body sway was not affected by environmental conditions. By using an air conditioner, we adjusted the room temperature to 25 °C and kept the room dark. All subjects were tested from 10 a.m. to 5 p.m. in the room. The subjects wore an HMD on which 2 kinds of images were presented in a random order
　　(Fig. 3a): (I) a static visual target (circle) whose diameter was 3 cm; (II) a conventional 3D movie that shows a sphere approaching and moving away from subjects irregularly (Fig. 3b). Resting state could be measured during the presentation of (I), and 3-D effects are evaluated by comparing with the resting state. 2.4 Procedure
　　The subjects stood without moving on the detection stand of a stabilometer in the Romberg posture (Fig. 4) for 1 min before the sway was recorded in order to measure stable time series. Each sway of the COP was then recorded for 120 s at a sampling frequency of 20 Hz during the measurement, while 3 dimensional head acceleration was simultaneously recorded by the active tracer at 50 Hz; subjects were instructed to maintain the Romberg posture for the first 60 s and a wide stance(with the midlines of heels 20 cm apart) for the next 60 s. The subjects viewed one of the images, i.e., (I) or (II), on the HMD from the beginning until the end. The SSQ was filled before and after stabilometry. An interval of
　　five minutes was set between above-mentioned measurements while subjects viewed images (I) and(II) so that previous one does not affect subjects’biosystems.
　　We calculated the total locus length (L), the area of sway (A), and the total locus length per unit area (L/A) that are commonly used in the clinical field as indices of stabilograms [21]. In addition, new quantification indices termed sparse density (SPD S2) and total locus length of chain (Chain) [22] were also estimated. The two-way analysis of variance (ANOVA) with repeated measures was employed for each index using a 0.05 criterion of statistical significance. We set posture(Romberg posture or standing posture with their feet wide apart) and images (I or II) as factors of fixed-effect model.
　　3. Results
　　The subjective SSQ data was obtained before and after the both measurements. According to the SSQ data obtained after the measurement while subjects viewed 3-D movie (II), scores for SSQ-N (nausea), SSQ-OD (eyestrain), SSQ-D (disorientation), and SSQ-TS (total score) were 11.4 ± 3.7, 18.2 ± 4.1, 23.7 ±8.8, and 19.8 ± 5.3 (mean ± SD), respectively. On the other hand, these SSQ subscores were almost zero in the other conditions.
　　3.1 Subjective Findings
　　Typical stabilograms are shown in Fig. 5. In these figures, the vertical axis shows the anterior and posterior movements of the COP, and the horizontal axis shows the right and left movements of the COP. The amplitudes of the sway that were observed during exposure to the movies (Figs. 5c-5d) were larger than those of the control sway (Figs. 5a-5b). Although a high density of COP was observed in the stabilograms(Figs. 5a-5b), the density decreased in stabilograms during exposure to the stereoscopic movie (Figs. 5c-5d). Furthermore, stabilograms measured in an open leg posture with the midlines of heels 20 cm apart(Figs. 5b, 5d) were compared with those measured in
　　the Romberg posture (Figs. 5a, 5c). COP was not isotropically dispersed but was characterized by considerable movement in the anterior-posterior (y) direction (Figs. 5b, 5d). Although this trend is seen in Fig. 5d, the diffusion of COP was larger in the lateral(x) direction and had spread to the extent that it was equivalent to the control stabilograms (Fig. 5a).
　　In addition to measuring the center of gravity, we simultaneously recorded simple moving average of the head acceleration at 20 Hz (Fig. 6). Peaks of a component in stabilograms seemed to be synchronized with maximal points of sequence in the head acceleration.
　　3.2 Statistics on Stabilograms
　　According to the two-way ANOVA with repeated measures, there was no interaction between the factors of posture and images (Table 1). For the SPD (S2), the
　　total locus length (L), and the area of sway (A) (Fig. 7), there were main effects in response to both factors (p < 0.01). In order to compare levels of the images or the postures, we also employed one-way ANOVA with repeated measures for each index. For any index, significant difference was observed between Romberg posture and standing posture with their feet wide apart
　　(p < 0.01). According to one-way ANOVA for SPD, L, and A, stabilograms of the resting state was significantly different from those observed when subjects viewed the 3-D movie (p < 0.01). However, there was no significance between images (I) and (II) for the other indices.
　　4. Discussion
　　Sickness symptoms seemed to appear with the exposure to the stereoscopic images in accordance with the result of the SSQ although there were large individual differences. Regarding to the body sway, the ANOVAs revealed that there were main effects in response to both factors (Romberg posture or standing posture with their feet wide apart) and images (I or II)) for the SPD (S2), the total locus length (L), and the area of sway (A). We also employed non-parametric statistics (Friedman test) whose result did not drastically changed. For the locus length per unit area(L/A) in addition to the above-mentioned indices, significant difference was also observed between images (I) and (II) (p < 0.05). The VIMS could be detected by these indices for stabilograms. Moreover, multiple comparisons (Tukey’s method) were employed as shown in Fig. 7. The only index is the SPD, which could obtain the statistical difference between stabilograms of the resting state and those observed when subjects viewed the 3-D movie (p < 0.01) independently of the postures.
　　The amount of body sway during a wide stance was less than that during the Romberg posture (Fig. 7). Subjects in this study must have a tolerance to motion sickness in accordance with Scibora, et al. (2007). These subjects were affected by viewing the 3-D movie(II), and their VIMS could be detected by calculating the SPD as an index of stabilograms.
　　In this study, we employed simultaneous recording of the center of gravity (Fig. 5) with the head acceleration during the exposure to a 2-D image and a 3-D movie(Fig. 6). Spikes of the head acceleration seem to be corresponding to the body sway (two black arrows).
　　We herein discuss quantitative interaction between the body sway and the head posture. Labyrinthine reflex plays an important part of the transfer system controlling the head posture (Fig. 1). When the subjects stood with the Romberg posture, transfer function analysis was implemented with the head acceleration(input) and the body sway (output); Based on a theorem(Winner-Khinchine):
　　WX f???????? (3) we can easily estimate a power spectrum xxW . On the right-hand side of Eq. (3), x? expresses the standard deviation and xx????? means the Fourier transform of the auto-correlation function with respect to the signal x(t) [23]. We herein estimate the transfer function that controls the sway; When subjects stood with their feet close together (Romberg posture), the coherence function between the head acceleration x(i) and the movement of the centre of gravity y(j) was estimated as
　　(4) where i and j expressed the component (1: lateral, 2: anterior/posterior). By using the Fast Fourier transform algorithm, power spectrums Wx(i)x(i), Wy(j)y(j) were estimated. On the basis of Eq. (3), we calculated cross spectrums Wx(i)y(j). The coherence means an index for the degree of the linear correlation between input and output signals (0 ≤ coh ≤1). There exists a completely linear correlation between these signals when coh = 1. In this study, we assumed that a linear system intervenes between the head and the body sway only if coh ≥ 0.12 (significant correlation coefficient for N = 512, p < 0.01). Moreover, we estimated the transfer function as follows:
　　 and the transfer function gain (TFG) |H(f)|.
　　We estimated the coherence function (4), i.e., cohx(1)y(1)(f), cohx(1)y(2)(f), cohx(2)y(1)(f) and cohx(2)y(2)(f)
　　Recently, a novel 3-D video construction method, Power 3D method [26] (Nishihira & Tahara, 2001) has been developed to prevent video-induced motion sickness. Humans perceive 3-D objects by simultaneous vergence and accommodation of the lens, but stereoscopic videos generally consist of the unnatural images perceived along a fixed optical axis of lens, negating such vergence and accommodation. In other words, the images are composed of photographs taken by two cameras or by computer graphics (CG). Camera axes are fixed and crossed at the point of the virtual image at which the creator expects viewers to gaze. That is, viewers would suffer from finding the anomalous vergence if they looked at the other elements in a stereoscopic flame. Thus, Power 3D method sets each camera axis as well as human beings that change the vergence angle corresponding to the visual distance of subjects for photography. Moreover, camera axes are also set to be parallel as well as natural binocular vision in order to construct the background. These elements of far/near visions are superimposed on the flame. Viewers might not feel a sense of incongruity if they gaze at any elements in the flame. We assume that stereoscopic images that are prepared using the Power 3D method reduce the inconsistency not only between vergence and accommodation but also experienced and actual senses. We have already reported some effects of Power 3D movie on the body sway [27-28]. The SPD was useful to compare the effects of 3-D movies as we mentioned in this study. In the next step, the above-mentioned assumption will be examined on the basis of simultaneous measurements of the vergence and the accommodation.
　　5. Conclusions
　　Visually induced motion sickness (VIMS) is caused by sensory conflict, the disagreement between vergence and visual accommodation while observing a 3-D movie. In order to evaluate the VIMS, we employed simultaneous recording of the center of gravity with the head acceleration during the exposure to a 2-D image and a 3-D movie in this study. The VIMS of subjects
　　that have a tolerance to motion sickness could be detected especially by calculating our SPD as an index of stabilograms. In addition to the analysis of stabilograms, we discussed that the anterior/posterior head acceleration could affect the lateral body sway during the VIMS caused by the 3-D movie.
　　 References
　　[1] T. Okawa, T. Tokita, Y. Shibata, T. Ogawa, H. Miyata, Stabilometry-significance of locus length per unit area(L/A) in patients with equilibrium disturbances, Equilibrium Res 55 (3) (1995) 283-293.
　　[2] K. Kaga, Memaino Kouzo: Structure of Vertigo, Kanehara, Tokyo, 1992, pp. 23-26, pp. 95-100.
　　[3] S. Morioka, F. Yagi, Influence of perceptual learning on standing posture balance: repeated training for hardness discrimination of foot sole, Gait and Posture 20 (2004) 36-40.
　　[4] K. Fujiwara, H. Toyama, Analysis of dynamic balance and its training effect—focusing on fall problem of elder persons, Bulletin of the Physical Fitness Research Institute 83 (1993) 123-134.
　　[5] T.A. Stoffregen, L.J. Hettinger, M.W. Haas, M.M. Roe, L.J. Smart, Postural instability and motion sickness in a fixed-base flight simulator, Human Factors 42 (2000) 458-469.
　　[6] G.E. Riccio, T.A. Stoffregen, An ecological theory of motion sickness and postural instability, Ecological Physiology 3 (3) (1991) 195-240.
　　[7] C. Oman, A heuristic mathematical model for the dynamics of sensory conflict and motion sickness, Acta Otolaryngologica Supplement 392 (1982) 1-44.
　　[8] J. Reason, Motion sickness adaptation: a neural mismatch model, J. Royal Soc. Med. 71 (1978) 819-829.
　　[9] T.A. Stoffregen, L.J. Smart, B.J. Bardy, R.J. Pagulayan, Postural stabilization of looking, Journal of Experimental Psychology: Human Perception and Performance 25 (1999) 1641-1658.
　　[10] H. Takada, K. Fujikake, M. Miyao, Y. Matsuura, Indices to detect visually induced motion sickness using stabilometry, in: Proc. VIMS2007, pp. 178-183.
　　[11] T. Hatada, Nikkei electronics 444 (1988) 205-223.
　　[12] R.S. Kennedy, N.E. Lane, K.S. Berbaum, M.G. Lilienthal, A simulator sickness questionnaire (SSQ): a new method for quantifying simulator sickness, International J. Aviation Psychology 3 (1993) 203-220.
　　[13] S.R. Holomes, M.J. Griffin, Correlation between heart rate and the severity of motion sickness caused by optokinetic stimulation, J. Psychophysiology 15 (2001) 35-42.
　　[14] N. Himi, T. Koga, E. Nakamura, M. Kobashi, M. Yamane, K. Tsujioka, Differences in autonomic responses between subjects with and without nausea while watching an irregularly oscillating video, Autonomic Neuroscience: Basic and Clinical 116 (2004) 46-53.
　　[15] Y. Yokota, M. Aoki, K. Mizuta, Motion sickness susceptibility associated with visually induced postural instability and cardiac autonomic responses in healthy subjects, Acta Otolaryngologia 125 (2005) 280-285.
　　[16] L.M. Scibora, S. Villard, B. Bardy, T.A. Stoffregen, Wider stance reduces body sway and motion sickness, in: Proc. VIMS 2007, pp. 18-23.
　　[17] R.E. Baloh, V. Honrubia, Clinical Neurophysiology of the Vestibular System, 2nd ed., FA Davis, Philadeiphia, 1989.
　　[18] M.W. Morningstar, B.R. Pettibon, H. Schlappi, Trevor TV: reflex control of the spine and posture: a review of the literature from a chiropractic perspective, Chiropractic & Osteopathy 13 (2005) 13-16.
　　[19] M. Sakaguchi, K. Taguchi, T. Ixhiyama, K. Netsu, K. Sato, Relationship between head sway and center of foot pressure sway, Auris Nasus Larynx 22 (3) (1995) 151-157.
　　[20] T. Takeda, S. Izumi, K. Sagawa, On the correlation between the head movement and the movement of the center of gravity using HMD, in: Proceedings of the 1995 IEICE General Conference, 1995, p. 203.
　　[21] J. Suzuki, T. Matsunaga, K. Tokumatsu, K. Taguchi, Y. Watanabe, Q&A and a manual in stabilometry, Equilibrium Res 55 (1) (1996) 64-77.
　　[22] H. Takada, Y. Kitaoka, S. Ichikawa, M. Miyao, Physical meaning on geometrical index for stabilometly, Equilibrium Res 62 (3) (2003) 168-180.
　　[23] K. Kido, Digital Fourier Transform (II), Corona Publishing, Tokyo, 2007, pp. 68-102.
　　[24] H. Takada, T. Yamamoto, M. Miyao, T. Aoyama, M. Furuta, T. Shiozawa, Effect of a stereoscopic movie on the correlation between head acceleration and body sway, Lecture Notes in Computer Science 5622 (2009) 120-127.
　　[25] B.L. Day, A.S. Cauquil, L. Bartolomei, M.A. Pastor, I.N. Lyon, Human body-segment tilts induced by galvanic vestibular stimulation: a vestibularly driven balance protection mechanism, J. Physiol. 500 (1997) 661-672.
　　[26] T. Nishihara, H. Tahara, Apparatus for recovering eyesight utilizing stereoscopic video and method for displaying stereoscopic video, U.S. Patent, US7404693B2, 2001.
　　[27] H. Takada, K. Fujikake, M. Omori, S. Hasegawa, T. Watanabe, M. Miyao, Reduction of body sway can be evaluated by sparse density during exposure to movies on liquid crystal displays, in: Proceedings of International Federation for Medical and Biological Engineering(IFMBE), 2008, Vol. 23, pp. 987-991.
　　[28] K. Fujikake, M. Miyao, T. Watanabe, S. Hasegawa, M. Omori, H. Takada, Evaluation of body sway and the relevant dynamics while viewing a three-dimensional movie on a head-mounted display by using stabilograms, Lecture Notes in Computer Science 5622 (2009) 41-50.