Keywords

1 Introduction

The development of 3D technology has involved all aspects of life, such as medical, industrial, military, education, entertainment, film, and television. The application of 3D technology also brings a new problem-visual fatigue. Many scholars have carried out related research, such as: Wang [1] studied the effect of visual fatigue on 3D polarized displays based on medical eye symptoms. Yano [2] pointed visual fatigue was induced if the images were moved in depth according to a step pulse function even if images were displayed within the corresponding range of depth of focus. Kooi et al. [3] proposed that crosstalk and blur may be the factors that cause visual fatigue. Matsuura [4] proposed that the influential factors affecting the viewing conditions in 3D stereo vision are interpupillary distance, visual function, and viewers. Alhaag et al. [5] confirmed that 3D display at short distance is more likely to cause visual fatigue than 2D display, but 3D display at a long distance has a lesser effect on visual fatigue.

For the measurement of visual fatigue, the most extensive subjective measurement is a questionnaire. For example, Lambooij [6] gave a visual fatigue questionnaire recognized by most scholars, namely the questionnaire for visual symptoms of Sheedy; Li et al. [7] used EEG and ERP signals to measure 3D visual fatigue and concluded that binocular parallax has effect on 3D visual fatigue. Tam et al. [8] indicated that there is a significant difference in 3D tolerance among individuals, and it is still impossible to distinguish whether it is caused by the stereo or the individual. Park and Mun [9] mentioned that it is also good to use eye movement parameters and physiological signals as 3D visual fatigue evaluation index when measuring visual fatigue. Kim et al. [10] used ECG, GSR, and SKT to measure the degree of fatigue affected by watching 3D video, indicating that skin electrical response (GSR) and skin temperature (SKT) may also be used to assess the extent of 3D visual fatigue. Bruce [11] measured drivers’ fatigue by recording heart rate and skin power. Ashrant et al. [12] used physiological indicators such as heart rate, EEG, and infrared temperature to detect the physical fatigue of construction workers. Zhu et al. [13] extracted the cycle, period standard deviation, amplitude, amplitude standard deviation, and frequent yawns of respiratory signals as characteristic parameters for determining driving fatigue. Yu [14] proposed a model for judging the current fatigue level and design and apply it to the mobile terminal (mobile phone application). Xu [15] judged the driver fatigue based on physiological signals EEG and ECG (heart rate signal). Ye [16] used physiological signals to classify driving fatigue and established a driving fatigue evaluation model by collecting and analyzing EMG and ECG. Fu et al. [17] used the wireless measurement equipment of physiological signals to detect the driver's fatigue state, and the prediction results based on the Bayesian model and reached a good agreement with the subjective score results.

In the current research, many scholars use physiological signals to evaluate fatigue, but the evaluation of respiratory signals is mainly for driving fatigue, which is different from visual fatigue caused by 3D display. Therefore, it is necessary to further correlate the respiratory characteristic parameters with 3D visual fatigue.

2 Experiment

2.1 Subject Selection

Thirty subjects (15 males and 15 females, aged from 19 to 45) were selected. The subjects had naked eye or corrected vision of 5.0. They did not perform strenuous exercise or watch movies and mobile phones for a long time before the test.

2.2 Equipment Parameters

LG42LW4500 (resolution 1920 × 1080, non-flash (i.e., polarized) 3D technology, screen ratio 16:9) was used with the parameters of contrast 95, backlight 100, and brightness 50 (recommended 3D optimal brightness). The measured brightness value in the mode is 102.12 cd/m2, and the photometer is Photo Research PR-680 model in the USA).

2.3 Experiment Procedure

  1. 1.

    Fill in the participant information record form and survey questionnaires.

  2. 2.

    Wear physiological signal sensor and setting the viewing distance to 1.5 m.

  3. 3.

    Measure the physiological parameters in the initial state for 5 min in a quiet state.

  4. 4.

    Fill out a questionnaire every 15 min, with a total of 6 times.

  5. 5.

    Keep and back up the test data and export the data.

3 Data Analysis

3.1 Questionnaire Data Processing and Analysis

Subjective data obtained from subjective questionnaire surveys. The mean statistics of the total subjective fatigue scores are shown in Table 1. The line chart of the changes in scores over time is shown in Fig. 1. At 60 min, due to the limited backup battery, battery replacement and recalibration procedures will be performed, which will make the subjects have 90 s rest time, so 60 min to 75 min, the score may not rise but may be the reason for getting rest. This also shows that during the period of visual fatigue, a short rest period can make people recover to a certain extent.

Table 1 Statistics of questionnaire score
Full size table
Fig. 1
figure 1

Polyline for mean of fatigue score

Full size image

Correlation analysis showed that at a test level of 0.01, the correlation coefficient between the fatigue score of the questionnaire and the time spent watching 3D videos was 0.894, which was extremely strong relationship. The Shapiro–Wilk test method was used to test the fatigue score data for normality. The test results are shown in Table 2.

Table 2 Results of normality test for the questionnaire
Full size table

Table 2 shows that the data for 45 min and after watching the 3D video conform to the normal distribution, and the p-values of the previous data are less than 0.05, which do not conform to the normal distribution. The Wilkerson symbol rank test and the paired t-test were used to compare the different viewing times with those when not watching, and the differences were analyzed. The results are shown in Table 3.

Table 3 Pared comparison for fatigue score
Full size table

From the pairwise comparison in Table 3, it is known that the p-value of the baseline data compared with the data of  30 min, 45 min, 60 min, 75 min, and 90 min are all less than 0.05. It shows that with the increase of time, subjective visual fatigue has shown a significant increase. In addition, the comparison of the data of the adjacent two groups shows that there is significant difference in the comparison between 30–45 min and 45–60 min, which indicates that the subject has a stronger sense of fatigue during this process, especially in the process of 45–60 min, which also has statistically significant at the significance level of 0.01.

3.2 RESP Signal Analysis

3.2.1 Time Domain Analysis

Perform time domain analysis on the RESP signal, and analyze the average value of the RESP over a period, that is, the average value of the respiratory frequency.

First, the data was tested for abnormality. Under the criterion of z = 3, there were no abnormal values. Correlation analysis of respiratory frequency to time showed that the respiratory frequency was significantly negatively correlated with time, with a correlation coefficient of −0.784 and a p-value of 0.037.

Descriptive statistics of the mean respiratory frequency showed that the mean value decreased from 24.5 rpm to 19.5714 rpm, but there was no significant change in the entire data and time, as shown in Fig. 2. The normality of each group of data was tested, and all groups were normally distributed. The paired samples t-test was performed on the results. The test results are shown in Table 4. The results show that, under the significance level of 0.05, although the mean value of respiratory frequency is significantly lower than that of the non-watched 3D video when watching 90 min of 3D video, this change has not shown statistical significance. Therefore, visual fatigue under 3D display cannot be accurately measured with the time domain characteristics of respiratory indicators.

Fig. 2
figure 2

Mean of RESP value to time

Full size image
Table 4 Paired samples t-test results
Full size table

3.2.2 Frequency Domain Analysis

The frequency domain analysis indicators of the breathing index are power and peak, respectively. The two indexes represent the power of the breathing band and the maximum value of the breathing frequency, that is, breathing power and breathing peak. Descriptive statistics of the two indicators are shown in Table 5.

Table 5 Paired samples t-test results
Full size table

Correlation analysis between power and peak values and viewing time shows that the average and standard deviation in power are strongly correlated and extremely correlated with time, while all parameters in the peak index are extremely weakly correlated or unrelated. Therefore, only the power value is analyzed.

In the power indicator, it can be seen clearly that the average value and the standard deviation decrease gradually. The power value decreases from 538.6423%2 in the initial state to 166.24%2 in the 90 min.

The data was abnormal at 60 min, which is consistent with subjective measurement data. It shows visual fatigue sharply from 45 to 60 min. And after adjustment, it returns to the normal level at 75 min. Removing the data at 60 min, a curve was drawn for the mean of power and standard deviation as showed in Fig. 3. It shows a general trend of slow decline. This indicator also reflects the decline of the human's breathing power, and the breathing gradually slows down. Combined with the subjective questionnaire, it appears along with visual fatigue caused by 3D video viewing.

Fig. 3
figure 3

Power mean and standard deviation along with time

Full size image

After entering the 3D video viewing state, the power mean basically shows an obvious linear distribution. Pearson correlation between power mean and time is −0.803 with sig.(2-tailed) 0.030, while the stand deviation is −0.779 and 0.039, separately.

The trend lines are added separately to find that the relationships between the power mean (Pm) for the breathing frequency domain and the time t for viewing the 3D video is consistent with:

$$P_{\text{m}} = - 1.6075t + 306.95,\;R = 0.9587.$$

Relationships between the standard deviation of the power mean (Psd) and the time t for watching 3D video:

$$P_{{{\text{sd}}}} = - 1.4277t + 298.54,\;R = 0.9534.$$

It can be seen from the fitting degree of the two trend lines that their fitting degrees are both greater than 0.95, which indicates good fitting degree of the regression line to time. Correlation analysis was performed between the mean and standard deviation of power and subjective scores, the correlation coefficients were both −0.971, and the p-value was 0. Therefore, visual fatigue can be evaluated using power or standard deviation. And the 3D video viewing time (greater than 15 min) can be estimated based on the power value of the RESP indicator, which provides guidance for the application of future physiological indicators to monitor 3D visual fatigue.

In order to observe the regularity of the data distribution further, a normality test was performed on the power value. The results are shown in Table 6.

Table 6 Power indicator normally test
Full size table

Based on the normality test results, under the test level of 0.05, five groups of power indicators do not meet the normal distribution, and two groups of peak indicators do not meet the normal distribution. The initial time is compared with the value of each the remaining time periods, and the paired sample t-test and Wilkerson signed-rank test are used for pairwise analysis, respectively. The results are shown in Table 7. There are significant differences between the viewing time and the initial time, while the p-value does not show a significant change trend. Among the comparison of neighboring 15 min, there are significant differences between 45 and 60 min and between 60 and 70 min, which is consistent with the questionnaire analysis.

Table 7 Pair test of power mean
Full size table

4 Conclusions

  1. 1.

    Questionnaire survey showed that subjective visual fatigue increased significantly after watching 3D videos for 30–60 min, especially during the period of 45–60 min, subjective visual fatigue increased significantly.

  2. 2.

    The consistency of the power mean and standard deviation in the breathing frequency domain to subjective visual fatigue is 0.971. The power mean and standard deviation shows strong correlation with time, and the correlation coefficient is −0.779 and −0.803, respectively.

  3. 3.

    After watching the 3D video for 15 min, the relationships between the power mean (Pm) and standard deviation (Psd) in the breathing frequency domain and time t meets: Pm = −1.6075t + 306.95, R2 = 0.9587; Psd = −1.4277t + 298.54, R2 = 0.9534.

  4. 4.

    The increase in 3D visual fatigue leads to a decrease in respiratory frequency, but the average respiratory frequency does not reflect the working time of visual fatigue; there is no significant correlation between peak value and visual fatigue.