Keywords

1 Introduction

In recent years, people have more and more mental work along with the rapid development of the economy. When people are fatigued, they often have poor physical coordination. The brain’s dominance of limbs declines, and people have slow thinking and slow movements, so it is easy for people’s judging mistakes and operational errors which will led to accidents [1]. At present, common fatigue testing methods mainly include subjective sensory evaluation method, physiological parameter test method, biochemical test method, psychological test method, and a combination of several methods [2, 3]. For example, Niu [4] used physiological signals such as ECG signals when constructing driving fatigue recognition models. Wu et al. [5] used EEG signals in driving fatigue detection research. Zhao et al. [6] used a combination of subjective questionnaires and physiological data detection of the drivers to determine driving fatigue. Yang [7] studied the impact of VDT mental fatigue status on the physiological electrical signals, and the electrocardiographic signals, pulse signals, analysis of physiological signals, and temperature signals. Cai [8] demonstrated the influence of human pulse signals in the detection of visual fatigue when studying the effects of visual fatigue on human physiological electrical signals and believed that pulse signals could achieve objective detection of visual fatigue. Zhang [9] used eye movement equipment to measure the relationship between pilot fatigue and PERCLOS and blink frequency. Wang et al. [10] used the change of heart rate index and subjective fatigue to study the effect of noise on construction fatigue and measured the level of fatigue of the subjects with evaluation table. Regarding the relationship between fatigue work and physiological signals, Jiang [11] mentioned in the experimental study of fatigue on physiological electrical signals that human fatigue can be intuitively reflected by changes in physiological parameters. Compared with domestic scholars, foreign scholars like to conduct in-depth exploration in many fields from ergonomics, physical mechanics, and medical biology. For example, Trejo [12] evaluates and classifies mental fatigue based on EEG. Borghini [13] measures the neurophysiological signals of aircraft pilots and car pilots to assess mental workload, fatigue and lethargy with EEG index, which controls the operator for brain fatigue monitoring through interactions between brain regions.

At present, many scholars’ research on mental fatigue mainly focuses on driver’ fatigue, psychological fatigue, visual fatigue, etc. [14]. Few studies are only related to mental work fatigue, and most scholars are focusing on EEG signals in the study of fatigue research. Strictly speaking, these researches are just some branches of mental fatigue. For example, driving fatigue includes mental fatigue and fatigue caused by static work, etc., while they are not the same type of mental fatigue. This paper mainly deals with mental work, which reflects mental fatigue from the perspective of physiological signals.

2 Experiment

2.1 Subjects Selection

The subjects were selected before the test, and 30 subjects (15 male and 15 female), aged 20–45 years old were invited. All the subjects, without physical discomfort, mental work, exercise, video work, entertainment, or using drugs, and so on before the test, were in good mental conditions.

2.2 Experiment Equipment

Wireless and wearable sensors were used to obtain the physiological parameters of the subjects, including: electrocardiograph monitoring sensor was used for electrocardiograph (ECG), myoelectric sensor was used to get electromyogram (EMG), pulse rate sensor was used to obtain photoplethysmographic (PPG) pulse waveform, and respiratory sensor was used to get respire (RESP) signals. Microsoft camera was used to capture the process of experiment. Human-machine-environment synchronization platform system, which was running on a computer, was used to collect data of physiological signals.

2.3 Experiment Procedure

Before the test was performed, the subjects were asked to rest for 20 min to fully ensure the stability of the test physiological signal. The temperature of the environment was constant at 23 °C and the procedures were as shown in Fig. 1.

Fig. 1.
figure 1

Flow chart of the experiment

Full size image

In Fig. 1, the questionnaire was mainly used to determine that the participants were fitting for participating in the test in order to avoid the outer factors that may affect physiological signals before the experiment and in the procedure. The relaxed state before the test was defined as the “basic state” (BS), which was mainly used to obtain the basic physiological signal data of the subject in a stable state. The state after the 5-min continuous high-intensity calculation test was defined as the “mental working state 1” (MWS1). After the visual fatigue scale test, another 10 min continuous high-intensity calculation test was carried on, and this state was defined as “mental working state 2” (MWS2). After the third visual fatigue scale test, a 15-min high-intensity calculation test was performed again, and this state was defined as “Mental Working State 3” (MWS3). In the flowchart, the visual simulation scale, in which “0” represents no fatigue and “10” represents very fatigue, was shown in Fig. 2. It was mainly used to provide subjective assessment for the fatigue levels of the subjects.

Fig. 2.
figure 2

Visual analog scale

Full size image

3 Data Analysis

3.1 Analysis of Visual Analog Ruler Measurement

According to the scale drawn by the subjects after each test, 30 subjects, of which 28 participates’ fatigue levels continued to deepen, one person’s fatigue level change was small, and one person showed signs of fatigue weakening as the mental work time increased due to the individual differences of the subjects. In a conclusion, the levels of fatigue to most subjects were deepened along with the experiment.

3.2 Analysis of Physiological Signal Measurement

The human-machine-environment synchronization platform was used to extract the physiological signals of the various fatigue states of the subjects. Since the sitting operation was used during the experiment, only mental work was performed, the operating environment was quiet, and there were no external influence factors. Therefore, the amplitude of physiological signals could be studied for further research of physiological signals influence on mental fatigue.

Physiological Signal Amplitude Changes.

In order to observe the peak amplitude variation of each physiological signals, the difference between the maximum value and the minimum value is subjected to descriptive statistical analysis, and the results are shown in Fig. 3, 4, 5 and 6. It shows in Fig. 3 that the peak amplitude variation of ECG signal increases gradually with the deepening of the fatigue. In Fig. 4 and Fig. 5, the amplitude variations of the EMG and PPG peak amplitudes increase at first and then decrease, while in Fig. 6 the change of the RESP peak amplitude is relatively stable. It indicates that there may be differences in physiological signals under different levels of fatigue. However, whether this difference is statistically significant or not and whether it is related to the level of brain fatigue or not are all needs further analysis.

Fig. 3.
figure 3

Curve of ECG peak amplitude variation

Full size image
Fig. 4.
figure 4

Curve of EMG peak amplitude variation

Full size image
Fig. 5.
figure 5

Curve of PPG peak amplitude variation

Full size image
Fig. 6.
figure 6

Curve of RESP peak amplitude variation

Full size image

Differences in the Mean Amplitude of Physiological Signals.

Multiple independent sample tests are performed on physiological signals under different working conditions. The results of the significance (2-tailed test) of the physiological signals in the experiments are shown in Table 1. It demonstrates that three states in the EMG test results do not meet normal distribution. And one of RESP signals also does not conform to the normal distribution. For data with normal distribution, the sample paired t test is used to analyze the difference, and for data that does not meet the normal distribution, Wilcoxon sign rank test is used.

Table 1. Shapiro-Wilk test results
Full size table

a. ECG Paired Sample Test

The differences in ECG amplitudes are in line with the normal distribution. Paired sample t-tests are used to perform a paired test on the four groups of data and Table 2 is obtained. It shows that there is no significant difference between the data of all paired tests. That is, whether the mental work is in level of fatigue or not, the mean amplitude of ECG signals cannot be used to reflect mental fatigue.

Table 2. ECG paired sample t test
Full size table

b. EMG Wilcoxon Rank Test

As it is shown in Table 3, although there is no statistical significance at the test level of 0.05 for all paired samples, p value of BS & MWS1, BS & MWS2, and BS & MWS3, getting by Wilcoxon rank test, are decreasing gradually.

Table 3. EMG Wilcoxon Rank Test Statistics
Full size table

The decrease indicates that the differences in EMG signals gradually appear as the levels of mental fatigue increases. EMG responds better to changes in mental fatigue, but it still lacks sensitivity.

c. PPG Paired Sample t Test

According to the test results of the paired samples shown in Table 4, there is statistically significant for BS & MWS2, which indicated that mental fatigue has affected PPG. The p value of BS & MWS2, MWS1 & MWS2 are all show significant differences at the test level of 0.05, while BS & MWS1, MWS2 & MWS3 do not show significant differences. It demonstrates that the MWS2 may be in fatigue state. However, the test results of BS & MWS3 does not appear significant differences. Therefore, the mean amplitude of the PPG peak should not be used to evaluate the level of mental fatigue.

Table 4. PPG paired sample t test
Full size table

d. RESP Data Analysis

The RESP data distribution of BS, MWS1, and MWS2 obey the normal distribution, and the paired sample t test is used for analysis. The results are shown in Table 5. BS & MWS3, MWS1 & MWS3, and MWS2 & MWS3 are analyzed using Wilcoxon rank test, as shown in Table 6.

Table 5. Paired sample t test
Full size table
Table 6. Wilcoxon Rank Test Statistics
Full size table

From the results of data analysis shown in Table 5 and Table 6, it appears that the significance (2-tailed) of BS & MWS1, BS & MWS2, and BS & MWS3 are gradually increases with the decrease of p value. And there is significant difference between BS and MWS3, which has statistical significance. There is no significant difference on MWS2 & MWS3, while there is more difference on MWS1 & MWS3 because the level of fatigue in MWS3 is higher. It proves that the RESP signals show clear correlation and regularity with the level of mental fatigue. Hence, the RESP signal can be used to reflect the level of mental fatigue and detect mental fatigue.

The non-linear fitting of the mean value of RESP amplitude and the mental time of work yields the equation:

T = 0.0099 RESP3 – 4.5313 RESP2 + 694.44 RESP – 35526, R2 = 1

R2 = 1, which shows that the formula has highly corresponding with actual value, as shown in Fig. 7. Therefore, the formula, which can be used to calculate the level of mental work under different intensities by RESP amplitude, provides a very useful evaluation mode for the evaluation of mental work.

Fig. 7.
figure 7

Relationship between high-intensity mental work time and mean amplitude of RESP peaks

Full size image

4 Conclusions

  1. 1.

    As the level of mental fatigue gradually increases, the mean amplitude of the RESP peaks gradually increases. When the high-intensity mental work last 30 min, the mean amplitude of the RESP peaks shows a significant difference at the test level of 0.05 (0.017).

  2. 2.

    The difference of EMG (ECG the same) mean amplitude increase gradually with levels of mental fatigue deepening, but it does not show statistical significance at the test level of 0.05. There are no significant correlations between the mean amplitude of PPG signals and mental fatigue.

  3. 3.

    A non-linear fitting of the mean amplitude of the breathing peak RESP and the mental work time T is performed, and a cubic polynomial is obtained.

T = 0.0099 RESP3 – 4.5313 RESP2 + 694.44 RESP – 35526, R2 = 1

According to this formula, R2 = 1, the high-intensity mental work time can be evaluated by the mean amplitude of the breathing peaks.