EEG-Based Driving Fatigue Alleviation Using Multi-Acupoint Electrical Stimulation Combine Music Conditioning Method

: In the present work, we propose the multi-acupoint electrical stimulation (stimulating the Láogóng 12 point ( 劳宫 PC8) and acupuncture points on waist, shoulders, buttocks of the human body) combined with music 13 conditioning (MESCMC) to alleviate driving fatigue. In our study, the complexity of α and β rhythms of EEG of 14 the drivers, the relative power spectrum of θ and β , as well as two relative power spectrum ratio θ/β and ( θ + α )/( α + β ) 15 are used as fatigue features during driving. The features of the complexity, which can effectively reflect brain 16 activity information, were used to detect the change of driving fatigue over time. Combined with the traditional 17 relative power spectrum features, the changes in driving fatigue features were comprehensively analyzed. The 18 results show that the MESCMC method can effectively alleviate the mental fatigue of drivers. Besides, compared 19 with the single-acupoint electrical stimulation[only stimulating the Láogóng point ( 劳宫 PC8)] (SES) method, the 20 MESCMC method is more effective in relieving driving fatigue.The mitigation equipment is low in cost and 21 practical, and the MESCMC method is individualized and improves the universality of driving fatigue detection 22 and relieve, so will be practical to use in actual driving situations in the future. 23


24
Many factors cause traffic accidents, among which traffic accidents caused by driving fatigue account for a 25 considerable proportion [1,2]. Therefore, it is particularly important for safe driving to accurately and quickly 26 detect the fatigue state of drivers and relieve the fatigue state in time in the process of long-term driving. EEG 27 signals can accurately represent the fatigue states of people [3], which is an ideal method of objectivity and 28 accuracy among many detection methods, is regarded as the "gold standard" [4-7]. Ma et al proposed a feature 29 extraction method based on a deep learning model to obtain higher classification accuracy and efficiency in EEG- 30 based driving fatigue detection [8]. Ren et al proposed a two-level learning hierarchy RBF network that allows for 31 global optimization of the key network parameters, aims to enhance the accuracy and efficiency of the EEG-based 32 driving fatigue detection model [9]. Min et al using the multi-entropy fusion method analyzed the multi-channel 33 area EEG to effectively detect the fatigue state of drivers, which achieved high-precision and high-sensitivity 34 driving fatigue detection [10]. 35 Merely detecting driver fatigue can sometimes not reduce safety hazards, so it needs to be alleviated in time to 36 reduce the incidence of traffic accidents. At present, the methods to relieve driving fatigue mainly include reducing 37 work intensity, taking short breaks, drinking caffeinated beverages, drugs, etc. These methods may affect normal 38 work or harm the body. In recent years, a new method of relieving human fatigue called electrical stimulation has 39 been widely used. Due to the advantages of small side effects of electrical stimulation, it has recently begun to be 40 used to alleviate human physical and mental fatigue [11,12]. Tsay  listening to music on the driver when driving in a low-complexity traffic environment. The results show that 48 listening to music in a monotonous car following task will not affect driving, and can improve driving ability such 49 as alertness [15]. Yuan et al analyzed the α rhythm power spectrum of the subjects who listened to different rhythms 50 of music, focusing on the corresponding relationship between EEG and music rhythm. The research showed that 51 music rhythm is an important factor affecting the α rhythm [16]. Trumbo et al adopted song naming games to 52 relieve driving fatigue in monotonous environments. The research showed that using song naming games as an 53 alleviation countermeasure can effectively alleviate driving fatigue [17]. Guo et al analyzed whether listening to 54 relaxing music can help alleviate mental fatigue and to maintain performance after a continuous performance task.

55
The results show that listening to relaxing music can reduce mental fatigue when performing persistent cognitive 56 tasks[18]. Li et al analyzed the influence of music rhythm on driver fatigue and attention quality in a long-distance 57 monotonous highway environment. The study showed that middle-speed music was the best choice to relieve 58 fatigue and maintain long-distance driving attention [19]. 59 In this study, the MESCMC method was used to relieve the mental fatigue of the subjects. This method uses a 60 conductive cloth that is fixed on the steering wheel of the car as the stimulation electrode. In the MESCMC method, 61 the music player plays a mid-rhythm (70-100 BPM) song while the subject is receiving electrical stimulation. 62 Nowadays, listening to music has become a habit of many drivers. Compared with traditional electrical stimulation, 63 this method does not require patch electrodes and is convenient to use. The music adjustment method and electrical 64 stimulation are activated at the same time to relieve fatigue. The MESCMC method is individualized and improves 65 the universality of driving fatigue detection and relieve. This method will have broad application prospects in the 66 future. 67 The Emotiv, as a portable EEG acquisition device with a sampling rate of 128Hz, is very convenient to be used in 79 real driving conditions. Its sensor electrodes are placed on the corresponding scalp surfaces of human brain regions 80

Experiment
according to the international 10-20 system (14 channels=AF3, AF4, F3, F4, FC5, FC6, F7, F8, T7, T8, P7, P8, O1  81 and O2). In the experiment, the Emotiv device is chosen for EEG signals recording. Each subject collected seven 82 stages of EEG signals. The EEG signals recording for each stage lasts 5 minutes. Recordings are performed with 83 the ears mastoids (right and left) used as the common reference. In the experiment, it should be ensured that all 84 leads are in a normal connection state during recording EEG signals for each stage. In addition, the electrical 85 stimulation and music plays must be suspended during EEG recording. The whole experiment lasted three hours 86 (2:00 p.m.∼ 5:00 p.m.). In the experiment of the MESCMC method to relieve driving fatigue, the flowchart of this 87 method is shown in Fig.1, k is the adjustment factor, which is adjusted according to the individual's fatigue 88 resistance. Considering the convenience of electrode placement and the efficiency of alleviating fatigue, we tried to 89 use Multi-acupoint electrical stimulation to alleviate driver fatigue. The location of the Láogóng point (劳宫PC8) 90 and multi-acupoint is shown in Fig. 3  The experimental platform can be divided into four parts: car simulator, Emotiv, electric stimulator and music player. In the experiment, 96 the car simulator is used to simulate the real driving environment; Emotiv that an EEG acquisition device was used to collect EEG of the 97 subjects; Electric stimulator and music player are used to relieve driver fatigue. The following is a brief introduction to the components of the 98 experimental platform. Fig.2 shows the experimental set-up. In the experiment, a multifunctional electrical stimulator (KWD-808I) was used to stimulate the acupoint with 102 a current of 1∼3mA (Subject to the strength that the participant can bear and is more comfortable), the pulse 103 frequency of the electric stimulation is 1Hz, and the intermittent wave, triangle wave and continuous wave are 104 alternately stimulated. In the experiment, the intensity of electrical stimulation must be lower than the level of 105 nerve damage that causes the human cortex to stimulate. The actual electrical stimulator is shown in Fig. 2.

106
In the experiment, Amoi Mp3 is used as the car music player. The player has 5D surround sound effects and 107 BOX shocking external playback. It supports TF card playback mode to meet various listening needs. Fig. 2 shows 108 the experimental set-up. 109 In our study, the subjects were asked to drive at 70Km-90Km/h on the highway, and drive along the route that 110 was told before the experiment. In addition, the experiment chose a low-complexity traffic setting as the driving 111 environment, and the weather was sunny. The research includes three types of driving experiments, namely driving 112 in normal driving conditions, driving with MESCMC method to alleviate driving fatigue, and driving with SES 113 method to alleviate driving fatigue. Each subject was tested in three driving states. In the normal driving 114 experiment, the subjects drive without electrical stimulation and music regulation. In the experiment of using the 115 SES method to relieve driving fatigue, the subjects turned on the electric stimulator after normal driving to the third 116 stage, and do not turn on the music player. In the experiment of using the MESCMC method to relieve driving 117 fatigue, after the normal driving to the third stage, the MESCMC method (shown in Fig. 1.) is used to relieve 118 driving fatigue. In this study, medium-speed music(70∼100 BPM)that the subjects are interested in and have a 119 good effect on alleviating driving fatigue is selected. 120

121
In this paper, wavelet packet decomposition (four-layer decomposition of frequency bands) is used to extract 122 δ(0∼4Hz), θ(4∼8Hz), α(8∼12Hz) and β(12∼32Hz) rhythms of EEG signals. Then, the complexity of α and β 123 rhythms of EEG signals of the drivers, the relative power spectrum of θ and β, as well as two relative power 124 spectrum ratio θ/β and (θ+α)/(α+β) are used as fatigue characteristics during driving. Finally, the efficiency of the 125 MESCMC method and the SES method in alleviating driving fatigue is analyzed. The following sections describe 126 these methods in detail. 127

EEG signals preprocessing 128
Since EEG signals collected in the driving environment contain strong interference, EEG must be preprocessing. 129 The WPD can overcome the defect that the frequency resolution of wavelet transform will decrease with the 130 increase of signal frequency, and then conduct more detailed analysis of the signal, can adaptively select the 131 frequency band, and better reflect the essential characteristics of the signal [20]. In this study, the WPD was used 132 for the preprocessing of EEG signals. The schematic diagram of the WPD is shown in Fig. 3, which is a four-layer 133 WPD. according to the resolution level. As the decomposition proceeds, the resolution in the time domain decreases, 138 while the resolution in the frequency domain increases. The WPD formula is as: 139 Let f(t) denote the source signals, after wavelet packet decomposition, 2i sub-bands are obtained in the ith sub-140 level, so the original signals can be expressed as: 141 is the reconstructed signals of the WPD on the node (i,j) of the ith layer. According to 142 Parseval's theorem and formula (1), the energy spectrum of signals f(t) wavelet packet decomposition can be 143 calculated as: 144 where E i,j (t j ) is the frequency band energy decomposed by the f(t) wavelet packet to the node (i,j); x i,j (j=0,1,2,…,2i-145 1, k=0,1,2,…,m) is the amplitude of the discrete points of the reconstructed signals f i,j (t j ); m is the number of signals 146 sampling points. In this paper, according to the WPD algorithm, binary scale transformation is used to decompose 147 the resampled EEG signals into the fourth layer to obtain the low-frequency sub-band of the EEG, as shown in 148 Table 1. 149 In this study, the 4-layer decomposition is made using the WPD to extract δ(0∼4Hz), (4∼8Hz), (8∼12Hz) 151 and (12∼32Hz) rhythms. The δ rhythm(0∼4Hz) reconstructed using the sub-band (4,0), the rhythm(4∼8Hz) 152 reconstructed using sub-band (4,1), the rhythm(8∼12Hz) reconstructed using sub-band (4,2), and 153 rhythm(12∼32Hz) reconstructed using sub-bands (4,3), (4,4), (4,5), (4,6), and (4,7). The waveform of each 154 rhythm after reconstruction is shown in Fig. 4.

Driving Fatigue detection Indicators 158
This study uses the complexity feature to analyze the tested EEG signal (non-stationary signal), and combines the 159 traditional relative power spectrum feature to comprehensively analyze the driving fatigue change process to 160 improve the accuracy and reliability of driving fatigue detection. 161

The complexity 162
The research showed that the higher the LZC (Lempel-Ziv complexity, LZC) value, the higher the probability 163 of the emergence of the new model, it also shows that the dynamic behavior is more complex [21,22]. Therefore, 164 LZC can reflect the changes of physiological signals with the state of body fatigue. The Lempel-Ziv complexity 165 algorithm proposed by Lempel and Ziv to measure the increase of new patterns as the sequence length increases. 166 The algorithm characterizes the rate at which new patterns appear in a time series. The calculation method of C 0 167 complexity is as: 168 Define {f(x), k=0,1,2,…,N-1} as a time sequence of length N, then the corresponding Fourier transform 169 sequence is: 170 where is the imaginary unit. For the formula (3), define W N =e -2πi/N , then there is: 171 where j=0,1,2,…,N-1. Assuming that the mean square value of {F N (j),j=0,1,2,…,N-1} is: 172 Perform inverse Fourier transformation on sequence {~( ), =0,1,2 , -1 The C 0 complexity formula is: 175 (8) Assuming that c(n) is the C 0 complexity of sequence S(s 1 ,s 2 ,…, s n ), Lempel and Ziv have proved that when 176 n→∞, c(n) approaches the fixed value n/log l n, then the normalized formula is: 177 where l is the number of coarse-grained stages (in traditional binarization, l=2). 178

Relative power spectrum 179
The studies have shown that when the driver is fatigued, the fast-wave (α and β) activity gradually decreases, the 180 slow-wave (δ and θ) activity gradually increases. Besides, the characteristics of the EEG power spectrum can 181 intuitively reflect the energy changes of EEG in different rhythms [23]. In this paper, the relative power spectrum is 182 used as a feature to detect the driving fatigue state. Assuming that the autocorrelation function of the random signal 183 is known, the power spectral density function can be defined as: 184 where r(k)=E[x(n)x*(n+k)]; E represents mathematical expectation; * represents complex conjugate.

185
In the experiment, the EEG signals of each subject were collected 7 times. The relative power spectrum of the 186 rhythm of the EEG signals refers to the ratio of the power spectrum in each rhythm of the EEG signals to the total 187 power spectrum of the EEG signals (0～32Hz), as shown in formula (11): where P i is the relative power spectrum of i rhythm, i = α, β, θ, δ.

190
The study showed that when adults change from a normal state to a fatigued state, the slow-wave of EEG 191 gradually increases, and the fast-wave gradually decreases (Borghini et al., 2014). Therefore, the sum of the slow-192 wave θ and the fast wave α can be combined with the fast wave, the ratio of the sum of slow-wave θ and fast-wave 193 α to the sum of fast-wave β and α energy can be used as an index to judge the fatigue state, denoted as F, and the 194 formula is: 195 where E θ , E α and E β represent the energy of θ, α and β rhythms respectively.

196
In this experiment, as driving time increases, the changes in θ and β rhythms are more regular than the 197 changes in the δ and α rhythms. In this experiment, β wave is selected as fast-wave, and θ wave is slow-wave. The 198 formula for selecting the index F to judge the fatigue state is simplified as: 199 In this experiment, the relative power spectrum average value of the β rhythm P β is used to replace the energy 200 E β of the β rhythm in formula (11), and the relative power spectrum average value of the θ rhythm P θ is used to 201 replace the energy E θ of the θ rhythm. The formula (11) can be expressed as: 202 In section 4.2 of this article, the linear fitting method is adopted. The relative power spectrum ratios and the C 0 203 values of the subjects who used the SES method and the MESCMC method to relieve driving fatigue are calculated 204 respectively. The linear slope indexes of the fitting are compared, which compared the efficiency of the SES 205 method and the SESCMC method to relieve driving fatigue. 206

The Complexity 208
Studies have shown that the complexity of EEG signals reflects the amount of EEG information and can 209 reveal the laws of brain activity [25-27]. Lempel-Ziv complexity (LZC) relates the complexity of a particular 210 sequence to the gradual emergence of new patterns in a given sequence [28]. The smaller the LZC of the EEG signal sequence, the more fatigued the driver [29,30]. In this paper, the C 0 values of and rhythms of one subject 212 were calculated, as well as the variation of the C 0 values of the subjects as driving over time were analyzed. The 213 trends of the C 0 value are shown in Fig. 5. corresponding to the three (normal driving group, MESCMC method group and SES method group) decreased 216 from starting the experiment to the third stage, which means that the fatigue of the subjects will gradually deepen 217 over time. From the fourth to seventh stages of the experiment, the C 0 values corresponding to and sub-bands 218 of the subjects continued to decrease for the normal group. 219 From the third stage of the experiment, as shown in Fig. 5, the MESCMC method is used to relieve driving 220 fatigue for the MESCMC method relief group, the subjects received multi-acupoint electrical stimulation and music 221 relieve for the MESCMC method relief group. The C 0 values of and bands of the subjects were an upward 222 trend after the fourth stage. For the SES method relief group, as shown in Fig. 6, only the single-acupoint electrical 223 stimulator was turned on after the third stage. The subjects in the SES method relief group received single-acupoint 224 electrical stimulation. The C 0 values of the and sub-bands of the subjects also have an upward trend. The 225 increase of C 0 values indicates that the driving fatigue of the 10 subjects (MESCMC method relief group and SES 226 method relief group) is alleviated. Moreover, the C 0 values, which subjects driving in the MESCMC method relief 227 group, are always greater than that of the SES method relief group, which means that the MESCMC method is 228 more effective in relieving driving fatigue compared with the SES method. 229 Fig. 5(a) and Fig. 5(b) show the brain topographic maps of the and sub-band the complexity when one 230 subject is in the MESCMC method relieves driving fatigue, the SES method relieves driving fatigue and normal 231 driving. The high complexity of the brain nerve is shown by a red shaded area, as well as the low complexity of the 232 brain nerve is shown by a blue shaded area. Fig. 5(a) and Fig. 5(b) show that the complexity of brain regions of 233 and sub-bands of the subjects show a significant downward trend in the 4-7 stages of normal driving. The 234 complexity of brain regions of and sub-bands of the subjects show a significant upward trend for the 235 MESCMC method relief group and the SES method relief group during the 4-7 stages of driving, which means the 236 MESCMC method and the SES method can alleviate driving fatigue. In addition, the and sub-bands of the 237 subjects in the relief group using the MESCMC method are significantly higher the complexity than the SES 238 method relief group, which means that the MESCMC method is superior to the SES method to relieve driving 239 fatigue. 240 In our study, we analyzed the C 0 values of and sub-bands of the 10 subjects of the two experiments(the 241 MESCMC method relief group and the normal driving group) in the 4-7 stages of the experiment, the trends are 242 shown in Fig. 6. The conclusion is attributed to the fact that the multi-acupoint electric stimulator and the music player are 250 turned on after the third stage. Moreover, the C 0 values corresponding to and sub-bands of the 10 251 subjects driving in the MESCMC relief group, are always greater than the normal driving group, which 252 means that the MESCMC method can effectively alleviate driving fatigue.

253
Studies have shown that EEG can reflect the neural activity of the human body [31,32]. In our study, It can be concluded from Fig. 7 that there is a significant difference in the C 0 values of and sub-260 bands of the 10 subjects between the MESCMC method relief group and the normal driving (p<0.01).

261
From the fourth to seventh stages of the experiment, the C 0 values corresponding to and sub-bands of 262 the subjects continued to decrease for the normal group. While the C 0 values of the and bands of the 263 subjects who were arranged to participate in two experiments(the MESCMC method relieve group and 264 the SES method relieeve group) also were an upward trend after the fourth stage. It means that the 265 MESCMC method and the SES method can effectively alleviate driving fatigue. Moreover, the C 0 values 266 corresponding to and sub-bands of the 10 subjects driving in the MESCMC relief group, are always 267 greater than that of the SES relief group, which means that the MESCMC method is more effective in 268 relieving driving fatigue compared with the SES method. driving fatigue is obtained. Firstly, to make various data comparable, the data is normalized. Taking the 278 relative power spectrum index of θ sub-band of one subject as an example, the least-squares method is 279 used to fit the relative power spectrum value of θ sub-band after the MESCMC method is used to relieve 280 driving fatigue. The fitted line is shown in Fig. 8. The SPSS software was used for analysis, the fitting line of the relative power spectrum of the θ sub-284 band of the subjects in the fatigue relief stage was obtained: y=-0.192*x+0.795, F-test p=0.027, which 285 passed the significance test and accorded with the linear relationship. According to the same method, 286 linear fitting was performed on the driving fatigue alleviation stage indicators of the two experiments 287 (MESCMC method relief group and SES method relief group), and the slope of the fitting line is shown 288 in Table 2. 289

292
In our study, the relative power spectrum of the θ and β sub-bands of the subject, the C 0 values 293 corresponding to the and β sub-bands of subject and the relative power spectrum ratio(θ+α)/(β+α) is 294 used as the fatigue feature. Table 2 shows the slope value of the fitting line corresponding to the fatigue 295 feature after the MESCMC and the SES method is used to relieve driving fatigue. After the subjects 296 received the MESCMC method to relieve fatigue, the absolute value of the slope of the fitting line of the 297 fatigue feature was greater than the absolute value of the slope of the fitting line of the fatigue feature 298 after the SES method was used to relieve the driving fatigue. Through the significance test, it is in line 299 with the linear relationship. The result showed that the MESCMC method is more effective than the SES 300 method in alleviating driving fatigue. 301

302
Previous studies have shown that when the driver is fatigued, his reaction ability, cognitive ability and 303 operation ability are severely affected, the judgment accuracy is reduced, and more and more errors occur, 304 which poses a serious threat to safe driving [34,35]. Therefore, rapidly and accurately detect and relieve 305 driving fatigue is particularly important for driving safety. The study has shown that stimulating the 306 Láogóng point (劳宫 PC8) can effectively relieve the mental fatigue drivers during long-term driving [12].

307
The study has shown that music can relieve driving fatigue in a monotonous environment [36]. However, 308 the traditional single method of relieving fatigue does not meet actual driving needs. The MESCMC 309 method is to stimulate the multi-acupoint (the Láogóng point (劳宫 PC8) and acupuncture points on waist, 310 shoulders, buttocks of the human body) with the electric stimulator while the music player is playing the 311 music that the driver is interested in to relieve mental fatigue of the drivers. In our study, we propose the 312 multi-acupoint electrical stimulation combined with music conditioning (MESCMC) to alleviate driving 313 fatigue. 314

The Complexity and the Linear Fitting Method 315
Previous study indicated that there is a downward trend of the complexity as subjects' fatigue deepens 316 [35,36]. This result is consistent with our research. The C 0 value reflects the probability of a new pattern 317 in the brain, and indirectly reflects the driving fatigue states of the subjects. In our study, we analyzed the 318 complexity, relative power spectrum, relative power spectrum ratio and linear fitting slope of different 319 rhythms of EEG signals of the subjects, as well as used the four indicators to distinguish the fatigue state. 320 We compared the effects of the SES method and the MESCMC method in alleviating driving fatigue. As 321 shown in Fig. 6, the complexity of the subjects driving in normal mode also showed overall downward 322 trends. However, from the fourth stage to the seventh stage of the experiment, the C 0 values of the and 323 sub-bands of the subjects in the MESCMC relief group showed upward trends because the brain nerves 324 were continuously stimulated at successive driving stages, which keep the nerves in a relatively exciting 325 state. This means that the MESCMC method can effectively relieve driving fatigue. For two experiments 326 (MESCMC method relief group and SES method relief group), subjects driving in the MESCMC state 327 can stay awake more effectively than in the SES state. When the subject is driving in an awake state, the 328 C 0 value is greater than the C 0 value in a fatigued state. From the fourth stage to the seventh stage of the 329 experiment, the C 0 values of the and sub-bands of the subjects in the MESCMC relief group are 330 greater than the C 0 values of the and sub-bands of the subjects in the SES relief group, as shown in 331 Fig. 5 and Fig. 7. 332 The linear fitting method is used to fit the absolute value of the index slope from the 4-7 stages of the 333 experiment. The absolute value of the index fitting slope of the MESCMC method is greater than the 334 absolute value of the index fitting slope of the SES method, as shown in Table 2, which means the 335 MESCMC method is more efficient than the SES method in alleviating driving fatigue. Because the 336 acupoint electrical stimulation can dredge the body's meridians, allow blood to pass through, protect 337 internal organs, and relieve physical fatigue. In the MESCMC method, stimulating electrodes are 338 arranged on the seat corresponding to the shoulders, waists, and buttocks of the human body to relieve 339 body stiffness and numbness caused by long-term driving of the human body. Music can act on the limbic 340 system and brainstem network structure of the human brain through human hearing. Meantime, music can 341 regulate the cerebral cortex, so that visceral activities of the body and behaviors have a well-coordinated 342 response. Furthermore, music relieves boredom and fatigue during driving. The MESCMC method is 343 more effective than the single the SES method to relieve driving fatigue. Additionally, the mitigation 344 equipment is low in cost and practical, and the electrical stimulation equipment of the MESCMC method 345 used multi-acupoint electrical stimulation that can alleviate driving fatigue more effectively than 346 traditional methods for relieving fatigue and improve driving safety. The MESCMC method is 347 individualized and improves the universality of driving fatigue detection and relieve, so will be practical 348 to use in actual driving situations in the future. 349

Novel Findings of This Study 350
This study adopted the MESCMC method to relieve driving fatigue. In the MESCMC method, 351 stimulating electrodes are arranged on the seat corresponding to the shoulders, waists, and buttocks of the 352 human body to relieve body stiffness and numbness caused by long-term driving of the human body. Our 353 study shows that compared with the SES method, the MESCMC method can more effectively alleviate 354 driving fatigue. In future research related to human mental fatigue, researchers can try to use the 355 MESCMC method to relieve human mental fatigue. Additionally, listening to music during driving has 356 become a driving habit of many drivers. The method adopts simple equipment and low cost, and can 357 alleviate driving fatigue more effectively than traditional methods for relieving fatigue and improve 358 driving safety. The MESCMC method is individualized and improves the universality of driving fatigue 359 detection and relieve. This method of relieving driving fatigue has positive significance for practical 360 applications in the future. 361

Limitations and Future Research Lines 362
In our experiments, when using an electric stimulator, how much stimulation current and stimulation 363 frequency will be more effective in alleviating driving fatigue requires further research. As the driving 364 fatigue state changes, when playing music, it is necessary to further study what rhythm of personalized 365 music should be played in different fatigue states to achieve the purpose of personalized music 366 recommendation. Electrical stimulators and music playback equipment need to be further integrated. 367

368
In our study, the MESCMC method was proposed for alleviating driving fatigue. The features of the 369 complexity, which can effectively reflect brain activity information, were used to detect the change of 370 driving fatigue over time. Combined with the traditional relative power spectrum features, the changes in 371 driving fatigue features were comprehensively analyzed. The results show that the MESCMC method can 372 effectively alleviate the mental fatigue of drivers. Besides, compared with the SES method, the MESCMC 373 method is more effective in relieving driving fatigue. Moreover, the mitigation equipment is low in cost 374 and practical, and the electrical stimulation equipment of the MESCMC method used multi-acupoint 375 electrical stimulation that can alleviate driving fatigue more effectively than traditional methods for 376 relieving fatigue and improve driving safety. The MESCMC method is individualized and improves the 377 universality of driving fatigue detection and relieve, so will be practical to use in actual driving situations 378 in the future. 379  384 The authors declare that there is no conflict of interests regarding the publication of this paper.

388
Technology Bureau (20166012), Central Guidance on Local Science and Technology Development Fund 389 of Hebei Province (206Z0301G). 390

Ethics approval 391
The Ethics Committee at the Northeast Electric Power University Hospital endorsed the study protocol. 392 All procedures performed in studies involving human participants were in accordance with the ethical 393 standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration 394 and its later amendments or comparable ethical standards. 395 references