Estimation of Degree of Interest in Comics Using a Stabilometer and an Acceleration Sensor

Abstract

The current widespread usage of the Internet has made it possible to easily browse through various types of web content. However, owing to the large amount of available web content, it is difficult to recommend items that match the viewer’s preferences. Although existing recommendation systems can recommend content based on the viewer’s browsing history and previous purchases, there is still a lack of content relevance. Hence, a system is required that can quantitatively evaluate the viewer’s degree of interest by incorporating biometric information and thereby recommend the appropriate content. In previous studies, the viewer’s concentration was measured by employing an acceleration sensor on the back surface of a chair. However, the viewer’s posture cannot be estimated when the viewer does not lean against the backrest. Hence, in this study, we propose a method for estimating the degree of interest by employing a chair equipped with a body stabilometer on the seat and an acceleration sensor on the back. In this study, when the subject was leaning against the backrest, we determined the position of the center of gravity by employing a body stabilometer, and we acquired acceleration data by employing an acceleration sensor. Furthermore, we analyzed the movement vectors of the position of the center of gravity and the acceleration. Consequently, the vector angle was divided after every 15°, and the analysis was conducted by examining the vector magnitude in the angle. The obtained results indicate a positive correlation between the interest in each story and the vector magnitude. Therefore, it can be concluded that the degree of interest can be evaluated by incorporating the vector magnitude of the position of the center of gravity and the acceleration.

1 Introduction

The current widespread usage of the Internet and smartphones has made it is possible to easily browse through numerous types of web content. However, owing to the large amount of available web content, it is difficult to recommend items that match the viewer’s preferences. Although conventional recommendation systems can recommend content based on the viewer’s browsing history and previous purchases, there is still a lack of content relevance. Hence, a system is required that can quantitatively evaluate the viewer’s degree of interest by incorporating biometric information and thus recommend the appropriate content.

In a previous study, Peter Bull [1] stated that the motion associated with “a body leaning forward” indicates the viewer’s high level of interest, and the motion associated with “a body leaning backward” indicates the viewer’s lack of interest. Based on this observation, Iwai et al. [2] studied the feature that an object of interest attracts the viewer’s neck to the front, and they estimated the degree of interest on the basis of gaze behavior and body movements. By mounting a distance sensor on the object, they estimated the degree of interest by measuring the distance between the subject and the object and measuring the duration of staring at the object with a camera. However, in such a case, the external device must be attached to the object. In another study, Sakamoto et al. [3] estimated the degree of concentration while browsing through comics by employing two body sway meters, but they did not consider cases involving a backrest because they used a chair without a backrest. Furthermore, Okubo et al. [4] measured the viewer’s concentration by employing a chair equipped with an acceleration sensor. However, posture estimation was not performed when the subjects were not leaning on the backrest.

Therefore, the objective of this study is to analyze the viewer’s behavior when presented with comics and to thereby evaluate the viewer’s degree of interest. In this study, we propose a method for estimating the degree of interest by using a chair equipped with a stabilometer on the seat and an acceleration sensor on the backrest.

2 Methodology

The experimental equipment is depicted in Fig. 1. A body stabilometer (with sampling frequency of 100 Hz) was placed on the subject’s chair. In addition, a cushion was placed on the chair to ensure that the subject did not feel uncomfortable. A Wii remote control was placed on the chair’s backrest as an acceleration sensor. The subjects rested briefly for approximately five minutes after sitting and were subsequently asked to read comics on the computer in a free posture. The subjects randomly read seven types of comics on 150–190 pages. After reading, they were asked to rate their interest in the comics on a scale of 1–10. The participating subjects were two healthy adults.

Fig. 1.

Experimental equipment

3 Analysis Indices

In this study, the outer peripheral area of movement and the maximum average size of vector (ASV) were incorporated as analysis indices during sway measurement of the center of gravity.

3.1 Peripheral Area

We determined the central position for the center of gravity sway in the sampling section. The moving outer circumference was thereby divided into 120 × 3°, and the area connecting the farthest points in each division was defined as the outer peripheral area. The detailed method is described below with the corresponding equations.

Let $A_{1} \left( {x_{1} ,y_{1} } \right),\,A_{2} \left( {x_{2} ,y_{2} } \right),\, \ldots ,\,{\text{and}}\;A_{n} \left( {x_{n} ,y_{n} } \right)$ represent the coordinate points. The point at the center of the section can be corrected as a temporary origin by obtaining the average of $x$ and $y$ values in the section and subtracting it from the corresponding coordinate values. Let $tA_{1} \left( {tx_{1} , ty_{1} } \right),\; tA_{2} \left( {tx_{2} , ty_{2} } \right),\; \ldots ,\;{\text{and}}\;tA_{m} \left( {tx_{m} , ty_{m} } \right)$ represent the corrected coordinates. Then, $tx_{m}$ can be expressed as follows:

$$ tx_{m} = x_{m} - \frac{{\mathop \sum \nolimits_{k = 1}^{n} x_{k} }}{n} $$

The angle $\theta$ and distance $d$ from the temporary origin can be calculated as follows:

$$ \theta_{m} = \frac{180}{\pi } \times \tan^{ - 1} (ty_{m} /tx_{m} ) $$

$$ d_{m} = \sqrt {tx_{m}^{2} + ty_{m}^{2} } $$

Here, $\theta$ denotes the frequency, which is in the range of −180° to 180°. We determined the position in the section where $\theta$ was divided into 3° units and thereby determined the farthest point among them. Let $G_{1} \left( {mx_{1} ,my_{1} } \right),\;G_{2} \left( {mx_{2} ,my_{2} } \right),\; \ldots ,\;{\text{and}}\;G_{120} \left( {mx_{120} ,my_{120} } \right)$ denote the determined coordinates. Then, the outer peripheral area $s$ can be expressed as follows:

$$ s = \mathop \sum \limits_{n = 1}^{120} \frac{1}{2}\left| {mx_{n} \times my_{n + 1} - my_{n} \times mx_{n + 1} } \right| + \frac{1}{2}\left| {mx_{120} \times my_{1} - my_{120} \times mx_{1} } \right| $$

3.2 Maximum ASV

We examined the vector for each sampling time (start point = coordinates of 0.01 s ago, end point = coordinates of time point of sampling). The vectors were divided into 15° units, and the characteristics of the vectors included in that direction were analyzed. Two features, namely the maximum ASV included in each 15° unit and the number of vectors, were examined.

Let $A_{1} \left( {x_{1} ,y_{1} } \right),\;A_{2} \left( {x_{2} ,y_{2} } \right),\; \ldots ,\;{\text{and}}\;A_{n} \left( {x_{n} ,y_{n} } \right)$ represent the coordinate points. Furthermore, let $n$ denote the number of samples in the target section. Then, $ASV$ can be expressed as follows:

$$ ASV = \frac{{ \mathop \sum \nolimits_{k = 1}^{n - 1} \sqrt {\left( {x_{k + 1} - x_{k} } \right)^{2} + \left( {y_{k + 1} - y_{k} } \right)^{2} } }}{n} $$

4 Results and Conclusion

In this study, the maximum ASV was calculated for each comic that the subjects read. The correlation coefficient between the maximum ASV and the degree of interest was calculated by evaluating the degree of interest after browsing, and analysis was conducted for each subject.

Firstly, the time spent leaning on the backrest while reading was extracted from the data of the acceleration sensor. The time spent leaning on the backrest was short in each experiment because the subjects browsed through the comics on the computer. Hence, the analysis was performed without discriminating whether the subjects leaned on the backrest or not.

After calculating the peripheral area and the maximum ASV for each comic, the correlation coefficient between the maximum ASV and the degree of interest was estimated. In addition, the correlation coefficient between the peripheral area and the degree of interest was also calculated. The results corresponding to the maximum ASV are illustrated in Fig. 2 and Fig. 3. It can be observed that the maximum ASV of the subjects and the degree of interest indicate a high correlation, which suggests that the maximum ASV of the center of gravity increases as the subject becomes more interested while reading the comics. The results corresponding to the outer peripheral area are illustrated in Fig. 4 and Fig. 5. It can be observed that the correlation coefficient of subject-1 in Fig. 4 is positive but less than or equal to 0.5. Furthermore, in Fig. 5, it can be observed that that the correlation with degree of interest is quite low for subject-2. Therefore, it is can be concluded that it is not sufficient to estimate the degree of interest solely on the basis of the peripheral area. It can also be concluded that irrespective of whether the subject leans on the backrest or not, the degree of interest can be estimated by incorporating the maximum ASV of the center of gravity sway.

Fig. 2.

Correlation between subject-1’s maximum ASV and degree of interest (R = 0.94)

Fig. 3.

Correlation between subject-2’s maximum ASV and degree of interest (R = 0.79)

Fig. 4.

Correlation between subject-1’s peripheral area and degree of interest (R = 0.44)

Fig. 5.

Correlation between subject-2’s peripheral area and degree of interest (R = 0.32)

5 Future Work

In this study, we attempted to analyze the viewer’s behavior upon being presented with comics. Hence, we developed a method for estimating the viewer’s degree of interest by incorporating the changes in the viewer’s posture, and the viewer’s body sway was analyzed by employing a stabilometer and an acceleration sensor.

Dividing the object according to the presence or absence of the backrest may affect the calculation of the correlation coefficient because the time spent leaning on the backrest was very short in this study. Therefore, the correlation between the maximum ASV and the degree of interest was calculated irrespective of the presence or absence of the backrest. Furthermore, it can be concluded that the degree of interest can be estimated by incorporating the maximum ASV of the center of gravity sway.

In this study, we did not analyze the presence or absence of the backrest. Hence, as future work, we would like to examine the possibility of estimating the degree of interest by asking the subjects to browse through a paper manga and categorizing the subjects into cases wherein subjects lean on the backrest or not.