A brain fatigue detection method with high sensitivity and high resistance to electromagnetic interference

Abstract

The present invention discloses a highly sensitive and highly electromagnetically resistant brain fatigue detection method, which is based on the 80Hz Donchirp sound signal stimulation to induce the subject to produce auditory steady-state response evoked potentials, and uses the instantaneous frequency multi-scale entropy MSE-IFV algorithm to analyze this rhythmic signal, successfully distinguishing the cortical EEG signals of brain fatigue and normal wakefulness, and has higher sensitivity and anti-interference ability than the Alpha wave and full wave analyzed by the same algorithm. It provides a highly anti-interference and highly sensitive method for the detection of brain fatigue state, which is suitable for brain fatigue detection in complex environments.

Description

Technical Field

[0001] The present invention relates to the field of brain fatigue detection technology, and in particular to a brain fatigue detection method with high sensitivity and high resistance to electromagnetic interference. Background Technology

[0002] Brain fatigue is caused by long-term monotonous work or excessive mental activities, which cause a large amount of free radicals, lactic acid and other substances produced by brain cell metabolism to accumulate, blocking the brain's nutritional pathways, causing a decrease in blood oxygen content, resulting in the inhibition of brain cell activity, and a decline in brain function performance, which is directly manifested as a decrease in work ability and reaction speed.

[0003] Currently, brain fatigue detection is very important for certain special occupations. For example, pilots need to detect brain fatigue recovery before flying before issuing flight missions. Bus drivers need to detect brain fatigue before and after driving for a period of time to determine whether they are suitable to continue working. EEG, as a device that directly measures brain electrophysiological signals, has the advantage of directness and is considered the "gold standard" for brain fatigue.

[0004] However, there are two difficult problems in the quantitative measurement of brain fatigue based on EEG methods: First, it is easily interfered by electromagnetic radiation in the environment, so it needs to be carried out in a shielded room, which limits its scope of application. Second, it is easily interfered by the transient state of the measured object in emotions, consciousness, attention, etc., and the obtained fatigue index has a large variance. Therefore, the "gold standard" of brain fatigue has become a "gold standard" in actual application. SUMMARY OF THE INVENTION

[0005] In view of the above problems, the present invention aims to provide a brain fatigue detection method with high sensitivity and high resistance to electromagnetic interference. Based on the mechanism of auditory steady-state response, the method induces the subject to produce auditory steady-state response evoked potentials, and realizes brain fatigue detection through multi-scale entropy analysis of evoked potentials. In order to achieve the above purpose, the technical scheme adopted by the present invention is as follows:

[0006] A highly sensitive and highly electromagnetically resistant brain fatigue detection method, characterized in that it comprises the following steps:

[0007] Step 1: Build a sound stimulation device to record EEG data before and after fatigue without stimulation;

[0008] Step 2: Select the stimulation frequency and sound intensity of the stimulation signal to obtain a specific sound stimulation signal;

[0009] Step 3: Apply specific sound stimulation signals to the subject and record the corresponding EEG data under stimulation;

[0010] Step 4: Extract the instantaneous frequency of the auditory steady-state evoked potential related to the sound stimulus from the EEG data;

[0011] Step 5: Perform complexity analysis on the extracted instantaneous frequency of the auditory steady-state evoked potential to obtain the time series of the instantaneous frequency variability of the auditory steady-state evoked potential IFV-ASSR;

[0012] Step 6: Perform multi-scale entropy analysis on the IFV-ASSR time series to obtain the distribution curve of sample entropy at scales 1-20;

[0013] Step 7: Fit the multi-scale entropy curve with a fourth-degree polynomial to obtain the polynomial as f(x i ), where x i is the scale factor;

[0014] Step 8: Differentiate the polynomial f(x i ) and use it as the slope of each scale Use k i to calculate β i , and β i = 2k i + 1; Based on the combination of multi-scale entropy, obtain the new complexity C i at 20 scales, C i = E i * β i , where E i is the multi-scale entropy at each scale factor;

[0015] Step 9: Average the new complexities at 20 scales and use it as the brain basal state index

[0016] Step 10: Calculate the brain fatigue index FI of the measured object, and its calculation formula is:

[0017]

[0018] Among them, F base is the brain basal state index in the non-fatigued state; F fatigue is the brain basal state index in the fatigued state;

[0019] Step 11: Determine whether the calculated brain fatigue index is within the normal value range. If not, it is considered that the current measured object is in a brain fatigued state.

[0020] Furthermore, the specific steps of Step 2 are as follows:

[0021] Step 21: Generate a Don chirp sound signal with a fixed frequency according to the Don chirp sound signal formula with a fixed frequency, and the Don chirp sound signal formula includes:

[0022] 1) The Don chirp signal latency-frequency formula is:

[0023] τ = kf -d (1-1)

[0024] τ is the latency, f is the stimulation frequency, k = 0.0920, d = 0.4356

[0025] 2) Phase delay:

[0026]

[0027] 3) The group delay is defined as the negative of the slope of the phase delay change:

[0028]

[0029] 4) The Chirp signal latency-frequency formula is:

[0030] t g = kf -d = cω -d (1-4)

[0031] Here, c is a constant, c = k(2π) d

[0032]

[0033]

[0034] Step 22: Save the generated Don chirp sound signal with the corresponding frequency as a wav format audio file.

[0035] Furthermore, the specific steps of Step 4 are:

[0036] Step 41: Extract the rhythm components from the EEG signal through the MSE-IFV algorithm. When extracting the rhythm signal, use a 4th-order Butterworth band-pass filter to filter out the signal with the corresponding frequency, and eliminate the phase distortion that appears after the Butterworth band-pass filtering through a zero-phase filter;

[0037] Step 42: Perform a Hilbert transform on the rhythm signal in Step 41 first to obtain the instantaneous frequency variability of the rhythm signal.

[0038] Furthermore, the specific steps for complexity analysis in Step 5 are:

[0039] Step 51: Normalize the complexity of the electroencephalogram (EEG) signal and convert it into the change rate of the cortical EEG signal before and after brain fatigue:

[0040]

[0041] where MSE fatigue represents the multi-scale entropy after fatigue, and MSE normal represents the multi-scale entropy when not fatigued;

[0042] Step 52: Take the data change rate of 10 s and perform 8 times of averaging:

[0043]

[0044] Step 53: Calculate the standard deviation based on Rate and :

[0045]

[0046] where n represents the number of times of averaging;

[0047] Step 54: Calculate the standard error of the sample mean SEM based on the standard deviation:

[0048]

[0049] where N is the number of subjects.

[0050] The beneficial effects of the invention are as follows:

[0051] First, the traditional method is to analyze spontaneous EEG (such as multi-scale entropy). The spontaneous potential of the human brain is divided into 4 frequency bands of δ, θ, α, and β according to the frequency passband. Each frequency band reflects various different states of the brain (concentration, emotion, thinking activity, fatigue, etc.). The existing detection of fatigue based on spontaneous EEG will inevitably be affected by non-fatigue factors. To address this drawback, the present application uses the preferred auditory steady-state evoked potential to avoid the range of δ, θ, α, and β spontaneous EEG, improving the sensitivity. Experiments show that the sensitivity is increased to 13.885%.

[0052] Second, the traditional method of recording EEG is easily affected by electromagnetic interference in the environment. The present invention can detect the high-interference frequency bands in the environment, select appropriate sound stimulation frequencies to avoid interference, and improve the signal-to-noise ratio. The pre-experiment results are improved to 5 times.

[0053] Thirdly, for the first time, the complexity algorithm is applied to the analysis of auditory steady-state evoked potentials. The traditional complexity algorithm (multiscale entropy) is improved, and a calculation formula for the brain fatigue index is proposed. The brain fatigue state of the person being tested can be simply and quickly detected through the obtained brain fatigue index. Description of the Drawings

[0054] Figure 1 It is the flowchart of the MSE-IFV algorithm in the present invention;

[0055] Figure 2 It is the Don chirp sound signal of different frequencies generated in the present invention;

[0056] Figure 3 It is the selection of the sampling rate;

[0057] Figures 4(a)-(b) show the changes in the complexity of electroencephalogram rhythm signals before and after the 2-hour pentagon flight mission in the present invention;

[0058] Figure 5 (a)- Figure 5 (d) are respectively the power spectrum diagrams of the resting-state electroencephalogram signals of the subjects in the shielding room and the cell room;

[0059] Figures 6(a) and 6(b) are respectively the complexity results of each rhythm component in the electroencephalogram signals obtained in the electromagnetic shielding room and the cell room;

[0060] Figures 7(a) and 7(b) are respectively the complexity results of each rhythm component in the electroencephalogram signals obtained in the electromagnetic shielding room and the cell room in the blinking and closing eyes states;

[0061] Figure 8 (a)-(e) is the flowchart of the flight fatigue experiment sequence;

[0062] Figure 9 (a)-(d) is the average change rate of the cortical electroencephalogram signals of the subjects after 2 hours of flight fatigue experiment;

[0063] Figure 10 (a)-(b) is the average change rate of the complexity before and after brain fatigue of 10 subjects in the embodiment;

[0064] Figure 11 It is the technical framework diagram of the present invention. Detailed Implementation Manner

[0065] In order to enable ordinary technicians in the art to better understand the technical solution of the present invention, the technical solution of the present invention will be further described below in conjunction with the drawings and embodiments.

[0066] Refer to the attached Figure 1 、 11A highly sensitive and highly anti-electromagnetic interference brain fatigue detection method is shown as follows, including the following steps:

[0067] Step 1: Build a sound stimulation device for playing audio files, design a brain fatigue induction experiment paradigm, and record the electroencephalogram data after brain fatigue without stimulation.

[0068] The experiment paradigm refers to that the subject conducts a 2-hour simulated flight fatigue experiment + real-time numerical calculation, and records the resting electroencephalogram data of the control group (without auditory evoked stimulation) and during auditory evoked stimulation before and after the flight fatigue experiment respectively.

[0069] Step 2: Select the sound intensity and stimulation frequency of the stimulation signal.

[0070] The auditory nerve has an inherent frequency. Giving a specific stimulus can induce certain fixed-frequency rhythm signals, namely auditory steady-state response evoked potentials. Auditory steady-state response evoked potentials are a kind of highly anti-interference evoked potentials. According to Fourier theory, by stimulating with a fixed-frequency Don chirp sound signal, the auditory steady-state response evoked potential can be obtained; and the fixed-frequency Don chirp sound signal can be generated by superimposing the pure tone signal of this frequency and the pure tone signal that is an integer multiple of its stimulation frequency at a fixed delay. Moreover, the stimulation of the fixed-frequency Don chirp sound signal can be deduced by the following formula:

[0071] 1) The latency-frequency formula of the Don chirp signal is:

[0072] τ = kf -d (1 - 1)

[0073] τ is the latency, f is the stimulation frequency, k = 0.0920, d = 0.4356

[0074] 2) Phase delay:

[0075]

[0076] 3) Group delay is defined as the negative of the slope of the phase delay change:

[0077]

[0078] 4) The latency-frequency formula of the Chirp signal is:

[0079] t g = kf -d = cω -d (1 - 4)

[0080] Here c is a constant, c = k(2π) d

[0081]

[0082]

[0083] Based on the Donchirp signal latency-frequency parameters of the above phase delay formula, the relationship between frequency and delay can be determined. Set the stimulation frequency as f1, and the harmonic signal of its integer multiple frequency is N*f1, where N is an integer. Select harmonics in the frequency range of 100 - 8000 Hz to generate a group of cosine pure tone signals. According to the above formula, substitute the harmonic frequency into the phase delay formula to calculate the phase delay t of each group of cosine signals. p After shifting each group of signals according to the corresponding delay and then superimposing them, a Don chirp signal with frequency f is generated. Convert the above formula into programming language in Matlab 2018b, and use Matlab 2018b to generate a Don chirp signal with the corresponding frequency, and save it as a.wav format audio file, then a sound stimulation signal that can generate auditory steady-state response evoked potentials can be obtained.

[0084] Step 3: Apply the sound stimulation signal to the object to be measured and record its current electroencephalogram data.

[0085] Specifically, let the subject wear in-ear headphones and play the Don chirp sound signals of this series of frequencies in the corresponding.wav format audio file (such as Figure 2 the signals with frequencies of 40, 70, 80, and 90 Hz shown) for stimulation, with the sound pressure level of 80 dB, and collect the cortical electroencephalogram signals of the subject; from the power spectrum of the electroencephalogram signals, this series of sound stimulations can all cause the subject to generate auditory steady-state responses.

[0086] Step 4: Extract the instantaneous frequency of the auditory steady-state evoked potential related to the sound stimulation in the electroencephalogram.

[0087] According to Figure 1 as shown, after the MSE-IFV algorithm extracts the rhythm components in the electroencephalogram signal, perform a Hilbert transform on this rhythm first to obtain the instantaneous frequency variability of the rhythm signal, and then perform MSE analysis on this variation signal to obtain the complexity of the variation signal of the rhythm signal.

[0088] When extracting the rhythm components from the original cortical electroencephalogram signal, use a 4th-order Butterworth band-pass filter to filter out the signals of the corresponding frequencies, and use a zero-phase filter to eliminate the phase distortion that appears after Butterworth band-pass filtering; then perform a Hilbert transform on the extracted rhythm signal to obtain the instantaneous frequency variability of the signal; finally, perform MSE analysis on this variation signal.

[0089] Step 5: Conduct complexity analysis on the instantaneous frequency of the extracted auditory steady-state evoked potential to obtain the time series of the instantaneous frequency variability of the auditory steady-state evoked potential (abbreviated as IFV-ASSR).

[0090] Since the individual differences in EEG signals are relatively large and the instantaneous EEG signals are significantly affected by emotions, the complexities of the cortical EEG signals before and after brain fatigue cannot be directly compared. Instead, the complexities of the cortical EEG signals before and after brain fatigue need to be normalized first and transformed into the change rates of the cortical EEG signals before and after brain fatigue:

[0091]

[0092] Among them, MSE fatigue represents the multi-scale entropy after fatigue, and MSE normal represents the multi-scale entropy when not fatigued;

[0093] Secondly, obtain the data change rate for 80 s and take the data for 10 s to make 8 averages:

[0094]

[0095] Then calculate the standard deviation:

[0096]

[0097] Among them, n represents the number of averaging times, and here n = 8;

[0098] Finally, obtain the standard error of the sample mean (SEM):

[0099]

[0100] Among them, N is the number of subjects;

[0101] Next, quantify the degree of brain fatigue through and observe the dispersion degree of the data through SEM;

[0102] Specifically, execute the MSE-IFV algorithm and complexity normalization processing in Matlab 2018 to calculate Write into the Excel table for storage and calculate the standard error of ;

[0103] Step 6: Conduct multi-scale entropy analysis (MSE) on the IFV-ASSR time series to obtain the distribution curve of the sample entropy at scales 1 - 20;

[0104] Step 7: Fit the multi-scale entropy curve with a fourth-degree polynomial to obtain the polynomial as f(x i ), where xi is the scale factor;

[0105] Step 8: Differentiate the polynomial and use it as the slope for each scale Utilize k i Calculate β i , β i = 2k i + 1, and obtain the complexity C of 20 new scales on the basis of combining multi-scale entropy i , C i = E i *β i , where E i is the multi-scale entropy on each scale factor;

[0106] Step 9: Average the new complexities on the 20 scales obtained and use the average data as the brain basic state index:

[0107]

[0108] Step 10: Calculate the brain fatigue index of the object under test. The calculation formula for the brain fatigue index FI is:

[0109] The brain fatigue index of an individual = (the brain basic state index in the non-fatigued state - the brain basic state index in the fatigued state) / the brain basic state index in the non-fatigued state, that is:

[0110]

[0111] where, F base is the brain basic state index in the non-fatigued state; F fatigue is the brain basic state index in the fatigued state;

[0112] Step 11: Determine whether the calculated brain fatigue index is within the normal value range. If not, it is considered that the current object under test is in a brain fatigued state.

[0113] Embodiment

[0114] To further verify the effectiveness and rationality of this evaluation method, a simulation experiment analysis is conducted.

[0115] I. Experimental instruments

[0116] Brain Products NSW316 16-channel wireless EEG cap system,

[0117] Thrustmaster T16000M flight joystick,

[0118] Thrustmaster T16000M throttle valve,

[0119] Huawei Mate20pro UD mobile phone,

[0120] Huawei CM33 in-ear wired earphones,

[0121] Matlab 2018b,

[0122] Start Neusen W software.

[0123] II. Experimental subjects

[0124] Multiple male and female subjects with normal hearing and no major diseases, aged 21 ± 1.2 years.

[0125] III. Experimental platform construction

[0126] 1. Construction of electroencephalogram (EEG) signal acquisition system

[0127] To reduce the discomfort of the subjects during the measurement of cortical EEG signals and provide higher activity, the Borui Kang NSW316 EEG cap system is used to collect cortical EEG signals. This system adopts the international 10 - 20 EEG lead system layout and has the ability to collect 16 - lead EEG signals. The EEG signals are picked up by electrodes, amplified and filtered analogically, then converted into digital signals through analog - to - digital conversion, and transmitted to a computer through a wireless network. The Start Neusen W software is used to complete the setting of EEG cap acquisition parameters and data recording. The sampling rate of the EEG cap is preset to 1000 Hz in Star Neusen W to ensure that the EEG signals are not distorted and can better reflect the instantaneous state, and the data is saved as a BDF file.

[0128] 2. Generation of sound stimulation signals

[0129] According to formulas (1 - 1)-(1 - 6), they are converted into programming language in Matlab 2018b, and a series of frequency Don chirp sound signals as shown in Figure 2 are generated by Matlab 2018b and saved as different wav - format audio files.

[0130] 3. Construction of flight simulation platform

[0131] The present invention uses flight simulation games and joystick operations to simulate the operation experience of pilots as much as possible, and at the same time gives the subjects a certain mental workload task to make them reach a state of mental fatigue through two hours of work.

[0132] X-PLANE is a flight simulation game that highly simulates real flight operations. Equipped with the Thrustmaster T16000M flight joystick and throttle valve, it realizes the simulation of basic flight operations. The subjects need to use the joystick to control the takeoff and landing of the aircraft and the control of various attitudes during flight. In this project, the commonly used international flight task of five-sided flight is adopted, allowing the subjects to take off the aircraft from the runway, fly around the runway clockwise or counterclockwise for one week, and then land back on the same runway.

[0133] IV. Experimental Methods

[0134] 1. Selection of the sampling rate of the MSE-IFV algorithm

[0135] Under the condition of a sampling rate of 1000Hz, simulation experiments were carried out to explore the influence of the sampling rate on the MSE-IFV algorithm. The original data with a sampling rate of 1000Hz was downsampled to 200, 400, 600, and 800Hz, and the influence of the algorithm at each sampling rate was examined. As shown by the simulation experiments Figure 3 below, the change is relatively large at 200Hz, and the entropy value change at 1000Hz is small, with better stability. The sampling rate of 1000Hz is more stable at scales 1-20 and has a better linear part. Therefore, a sampling rate of 1000Hz was adopted for the selection of the sampling rate of the MSE-IFV algorithm.

[0136] 2. Flight fatigue test under sound stimulation signals

[0137] As shown in Figures 4(a)-(b), when the subjects completed the two-hour five-sided flight task, the cortical electroencephalogram signals of the subjects were recorded when sitting quietly with eyes closed without sound stimulation and under the 60dB Don chirp sound stimulation. The subjects were made to wear in-ear headphones, and the Don chirp sound signals of different frequencies were played for sound stimulation with a sound pressure level of 60dB, and the cortical electroencephalogram signals of the subjects were collected. From the power spectrum of the electroencephalogram signals, this series of sound stimulations can all cause the subjects to produce auditory steady-state responses. Using the MSE-IFV algorithm to calculate the complexity of the Alpha wave and ASSR evoked potentials, from the perspective of complexity, after continuously repeating the five-sided flight task for two hours, the brain fatigue degree of the subjects has increased. Thus, it can be seen the stability and superiority of the ASSR wave. Although the alpha wave can also increase, it does not increase as much as the ASSR wave.

[0138] 3. Anti-interference ability test of ASSR evoked potentials under different electromagnetic environments

[0139] There is less electromagnetic interference in the electromagnetic shielding room. In the cell room, the instruments and equipment are concentrated and the electromagnetic environment is complex. The electromagnetic shielding room and the cell room are selected as two different test environments, and the self-control method is adopted. In the experimental group, the subjects are required to sit quietly with their eyes closed for two minutes in the cell room. Wear in-ear headphones and record the cortical electroencephalogram signals with Don chirp sound stimulation and without sound stimulation respectively. In the control group, in the electromagnetic shielding room, with the same two postures, wear in-ear headphones and record the cortical electroencephalogram signals with Don chirp sound stimulation and without sound stimulation respectively. Finally, compare and analyze the complexity of ASSR evoked potential, Alpha wave and full wave of the same subject in different environments.

[0140] 4. Anti-interference test of ASSR evoked potential in blinking and eyes-closed states

[0141] Similar to the above test, the self-control method is also adopted. The subjects are required to conduct the experiment in two physiological states: 1. In the experimental group, blink once every five seconds and sit quietly for two minutes; 2. In the control group, sit quietly with eyes closed for two minutes. Conduct a control experiment once in the electromagnetic shielding room and the cell room respectively. Finally, analyze and compare the complexity of ASSR evoked potential, Alpha wave and full wave of the same subject in the blinking and eyes-closed states in the same environment.

[0142] Collect electroencephalogram signals in the shielding room and the cell room respectively, and then use Fourier transform to obtain its power spectrum. The power spectrum diagrams of the resting-state electroencephalogram signals of the subjects in the shielding room and the cell room as shown in Figure 5 (a)-(d) can be obtained. The main electromagnetic interference in the environment is the 50Hz power frequency interference and its harmonics. The 40-90Hz Don chirp sound signals can all induce the auditory steady-state response evoked potential of the subjects. Among them, the auditory steady-state response evoked potential induced by the 90Hz Don chirp signal is mainly around 98Hz, close to the harmonic with a frequency of 100Hz. Therefore, the selection of the sound stimulation frequency should avoid the frequencies close to 50Hz and its harmonics. Finally, the 80Hz Don chirp sound signal is selected as the induced audio stimulus for the auditory steady-state response. Figures 6(a)-(b) show the complexity results of each rhythm component in the electroencephalogram signals obtained in the electromagnetic shielding room and the cell room respectively. Select the cortical electroencephalogram signal of the Pz lead, with a sampling rate of 1000Hz, a sampling length of 10 seconds each time, and a total of 8 times are taken to obtain an 80-second average result. It can be seen that for the same subject, on the premise of the same mental state, the complexity of its ASSR evoked potential is basically the same in the electromagnetic shielding room and the cell room, and from a statistical point of view, the two groups of data cannot be effectively distinguished. Under the condition of no sound stimulation, the complexity of the Alpha wave and the full wave of the same subject is significantly different in the electromagnetic shielding room and the cell room.

[0143] Figures 7(a)-(b) show the complexity results of each rhythm component in the EEG signals obtained in the electromagnetic shielding room and the cell room under blinking and closing eyes states, respectively. The cortical EEG signals of the Pz lead were selected, with a sampling rate of 1000 Hz, a sampling length of 10 seconds each time, and a total of 8 samplings were taken to obtain an 80-second average result. For the complexity of the two physiological states of sitting with eyes closed and sitting with blinking, the ASSR evoked potential showed high consistency, while the Alpha wave and the full wave showed differences about 5.6 times and 12.2 times higher than the ASSR evoked potential, respectively.

[0144] It can be seen from the above experimental results that the auditory steady-state response evoked potential formed in the cerebral cortex has higher anti-interference ability compared with the spontaneous Alpha wave and full-wave EEG signals of the human body. In different electromagnetic environments, the complexity performance of the auditory steady-state response evoked potential of the subjects in the same state shows high consistency. Even the complexity of the EEG signals measured in the non-electromagnetic shielding room environment can be close to that of the signals measured in the electromagnetic shielding room. In the tests of blinking and closing eyes, the auditory steady-state response evoked potential still showed higher consistency compared with the spontaneous Alpha wave and full-wave EEG signals, proving that the auditory steady-state response evoked potential can effectively reduce the influence of the electrophysiological signals generated by natural physiological activities such as human blinking on the cortical EEG signals.

[0145] The experimental results prove that the auditory steady-state response EEG signal analysis method has higher anti-interference ability and adaptability compared with the traditional methods of analyzing spontaneous EEG rhythm components and full-wave EEG analysis, and it is a means of detecting EEG signals that can serve in complex environments.

[0146] 6. Complexity test of EEG after brain fatigue

[0147] During the operation of the drone, many computing and input tasks other than flight control need to be faced. In the design of this experiment, in order to be as close as possible to the actual operation of the drone operator and make the subjects reach the state of brain fatigue as soon as possible within 2 hours, a mental arithmetic task was added to the flight driving task. Perform addition and subtraction within three digits, and require the subjects to complete the input task of mental arithmetic data according to a certain step sequence.

[0148] An mental arithmetic APP was developed through the APP designer in MATLAB. Two sets of addition and subtraction problems within three-digit numbers were designed in advance, with 30 problems in each set, arranged from easy to difficult. The subjects were required to complete one set of calculations every hour. After the experiment started, all the problems were printed in a table and given to the subjects. The subjects needed to perform a series of operations on the designed APP, such as clicking buttons in sequence, entering the calculated numbers, completing mental arithmetic, and entering the mental arithmetic results. Inside the APP, the mental arithmetic results of the subjects were automatically compared with the correct answers, with correct recorded as 1 and incorrect recorded as 0. The tic / toc function in MATLAB was used to be triggered synchronously when the subjects clicked the buttons to record the time required for the subjects to complete each problem. Finally, all the data was written into an EXCEL table for storage.

[0149] According to the experimental sequence shown in Figure 8 (a)-(e), the cortical electroencephalogram signals of the subjects were recorded before and after performing the 2-hour flight fatigue experiment. The MSE-IFV algorithm was run using Matlab 2018b to calculate the complexity of the auditory steady-state response evoked potential, Alpha wave, and full-wave electroencephalogram signals before and after brain fatigue. Finally, statistical analysis was completed in Excel and OriginPro 2021.

[0150] The statistical analysis of the obtained experimental data was implemented using OriginPro 2021 and Excel, and all were presented in the form of mean ± standard error of the sample mean (SEM). The general situation of the brain fatigue state detection of 20 subjects is shown in Table 1.

[0151] Table 1 General situation of the brain fatigue state detection of 20 subjects

[0152]

[0153]

[0154] Results are shown in Table 1: Among these 20 subjects, the ASSR evoked potential successfully detected the brain fatigue status of 18 subjects. There were significant differences in the ASSR evoked potential of these 18 subjects before and after brain fatigue, which could effectively distinguish between brain fatigue and normal waking state. The differences in the ASSR evoked potential of 2 male subjects before and after brain fatigue were too small to distinguish between brain fatigue and normal waking state, and the detection rate was 90% (SEM = 0.0089). For the Alpha wave, a decrease in complexity could be observed in 12 subjects at a high scale, but the degree of decrease was lower than that of the ASSR evoked potential, and the detection rate was 60% (SEM = 0.0425). In addition, the SEM of the average change rate of the Alpha wave was too large, indicating that the complexity of the Alpha wave fluctuated too much and the samples had a high degree of discreteness, and could not well represent the overall situation. For the full wave, only 1 male subject observed a decrease in complexity during brain fatigue at a high scale, and the change rates of the complexity of the full wave signals of other subjects fluctuated around 0.

[0155] Figure 9 (a)-(d) show the average change rates of the cortical electroencephalogram signals of 4 subjects after a 2-hour flight fatigue experiment. Regardless of the scale, the average change rate of the ASSR evoked potential remained stable below 0. The change rates of the Alpha wave of most subjects fluctuated above 0 at a low scale and could be stable below 0 at a high scale, and could distinguish between brain fatigue and normal waking state at a high scale. However, the average change rate of the Alpha wave before and after brain fatigue was lower than that of the ASSR evoked potential, and its SEM was too large, indicating that the data discreteness was too large and the samples were not sufficient to represent the whole. The change rate of the full wave always fluctuated around 0, and this fluctuation was independent of the scale, that is, regardless of which scale, the full wave could not show that there were significant differences in the complexity of the electroencephalogram signals before and after brain fatigue.

[0156] After the subjects underwent a 2-hour flight fatigue experiment, combining the average change rates before and after brain fatigue and the subjective feelings of the subjects, the conclusion was obtained with a scale of 20 as the standard: the complexity of the ASSR evoked potential decreased by about 15% per capita, the complexity of the Alpha wave decreased by 8% per capita, and the full wave did not have the ability to detect brain fatigue.

[0157] The change rates of the complexity before and after brain fatigue of 10 subjects were randomly selected from the 12 subjects who could distinguish between brain fatigue and normal waking state by both the ASSR evoked potential and the Alpha wave. After taking the average, we got Figure 10Results shown in (b): The average decline rate of Alpha waves at 20 scales is 0.95% (SEM = 0.0425); the average decline rate of ASSR evoked potentials at 20 scales is 13.89% (SEM = 0.0089); the rate of change of complexity of the full wave fluctuates around 0, and the brain fatigue state cannot be distinguished. From scale 1 to 20, the sensitivity of ASSR evoked potentials to brain fatigue and normal waking states is higher than that of Alpha waves, and the difference is more significant than that of Alpha waves. Taking the average of the complexity differences at scales 16 - 20, the average difference of ASSR evoked potentials is 4.191 times that of Alpha waves. At scale 20, the difference of ASSR evoked potentials is 2.874 times that of Alpha waves. In addition, when the scale is less than 10, the rate of change of complexity calculated from Alpha waves often fluctuates above 0 before and after brain fatigue, and the brain fatigue and normal waking states cannot be distinguished.

[0158] From Figure 10 It can be observed from (a)-(b) that at low scales or even single scales, ASSR evoked potentials can effectively distinguish between brain fatigue and normal waking states. This indicates that when applying the method proposed in this study to actual detection, there is no need to perform high-scale entropy calculations. Only low-scale entropy algorithms or even single-scale sample entropy algorithms can be used to detect the brain fatigue state. Reducing the calculation scale will greatly improve the operation speed, reduce the load of the hardware during operation, further improve the detection ability of the brain fatigue state, and make the detection more real-time and convenient.

[0159] From the results of the flight fatigue experiment on 20 people, Alpha waves and full-wave electroencephalogram signals have low sensitivity and poor universality in the detection of brain fatigue in the population. ASSR evoked potentials have high sensitivity and universality. Using ASSR evoked potentials can more effectively detect people with brain fatigue. Analyzing ASSR evoked potentials using the MSE-IFV algorithm is a stable and widely applicable method for detecting brain fatigue.

[0160] V. Experimental Conclusions

[0161] First, a comparative analysis of the MSE-IFV algorithm was performed on ASSR evoked potentials induced by Don chirp sound signals collected in an electromagnetic shielding room and a cell room and spontaneous electroencephalogram signals without sound stimulation. The experimental results show that in these two environments with significantly different electromagnetic environments, the complexity of ASSR evoked potentials shows high consistency compared with spontaneous full-wave electroencephalogram and Alpha waves.

[0162] Second, the MSE-IFV algorithm analysis results of ASSR evoked potentials, full-wave EEG, and Alpha waves without sound stimulation were compared and analyzed in the blink and closed-eye states. The difference in the complexity of ASSR evoked potentials between closed eyes and blinks was much smaller than that of Alpha waves and full waves.

[0163] Third, through a 2-hour flight fatigue experiment on 20 subjects, the complexity of EEG signals before and after brain fatigue was compared and analyzed based on the MSE-IFV algorithm. Brain fatigue and normal waking states were effectively distinguished by ASSR evoked potentials, with a detection rate of 90% (SEM = 0.0089); at scales higher than 10, the detection rate of Alpha waves was only 60% (SEM = 0.0425); the full wave could hardly effectively detect the brain fatigue state. Ten subjects were randomly selected from 12 subjects whose brain fatigue states could be distinguished by both ASSR evoked potentials and Alpha waves. After averaging the complexity change rates, it was found that the average decrease rate of Alpha waves at 20 scales was 0.95% (SEM = 0.0425); the average decrease rate of ASSR evoked potentials at 20 scales was 13.89% (SEM = 0.0089). By comparing and analyzing the complexity of ASSR evoked potentials before and after brain fatigue, with 20 scales as the standard, it was found that after a 2-hour flight fatigue experiment, the complexity of ASSR evoked potentials of 18 subjects decreased by approximately 15.65% on average (SEM = 0.0078).

[0164] Fourth, it was observed that ASSR evoked potentials already had the ability to distinguish brain fatigue and normal waking states at low scales or even single scales. In the subsequent application of this study to practice, a low-scale or single-scale sample entropy algorithm can be considered to improve the detection speed of the brain fatigue detection system, reduce the hardware operation load, and achieve real-time acquisition and real-time analysis as much as possible.

[0165] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification only illustrates the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A highly sensitive and highly resistant to electromagnetic interference brain fatigue detection method, characterized in that: The following steps are involved: Step 1: Build a sound stimulation device and record EEG data before and after fatigue without stimulation; Step 2: Select the stimulation frequency and sound intensity of the stimulation signal to obtain a specific sound stimulation signal; Step 3: Apply a specific sound stimulation signal to the subject and record the corresponding EEG data under the stimulation; Step 4: extract the instantaneous frequency of the auditory steady-state evoked potential associated with the sound stimulation in the EEG data; Step 5: Perform complexity analysis on the extracted instantaneous frequency of the auditory steady-state evoked potential to obtain the instantaneous frequency variability time series IFV-ASSR of the auditory steady-state evoked potential; Step 6: Perform multi-scale entropy analysis on the IFV-ASSR time series to obtain the distribution curve of sample entropy at scales 1-20; Step 7: Fit the multi-scale entropy curve with a quartic polynomial and obtain the polynomial f(x i ), where x i is the scale factor; Step 8: For the polynomial f(x i ) is differentiated and used as the slope of each scale Using k i Calculate β i , and β i =2k i +1; Based on the combination of multi-scale entropy, a new complexity C of 20 scales is obtained i , C i =E i *β i , where E i is the multi-scale entropy on each scale factor; Step 9: Average the new complexity at the 20 scales and use it as the basic brain state index Step 10: Calculate the brain fatigue index FI of the subject, the calculation formula is: Among them, F base F is the basic brain state index in non-fatigue state; fatigue It is the basic brain state index under fatigue state; Step 11: Determine whether the calculated brain fatigue index is within the normal value range. If not, it is considered that the current subject is in a brain fatigue state.

2. A brain fatigue detection method with high sensitivity and high resistance to electromagnetic interference as claimed in claim 1, characterized in that: The specific steps of step 2 are: Step 21: Generate a Don chirp sound signal with a fixed frequency according to a Don chirp sound signal formula with a fixed frequency, and the Don chirp sound signal formula includes: 1) The Don chirp signal latency-frequency formula is: τ=kf -d (1-1) τ is the latency, f is the stimulation frequency, k = 0.0920, d = 0.4356 2) Phase delay: 3) Group delay is defined as the negative of the slope of phase delay change: 4) The Chirp signal latency-frequency formula is: t g =kf -d =cω -d (1-4) Here c is a constant, c = k(2π) d Step 22: Save the generated Don chirp sound signal of the corresponding frequency as a wav format audio file.

3. A highly sensitive and highly electromagnetic interference resistant brain fatigue detection method as claimed in claim 1, characterized in that: The specific steps of step 4 are: Step 41: extract the rhythm component in the EEG signal by using the MSE-IFV algorithm, and use a 4th-order Butterworth bandpass filter to filter out the signal of the corresponding frequency when extracting the rhythm signal, and use a zero-phase filter to eliminate the phase distortion after the Butterworth bandpass filtering; Step 42: Perform a Hilbert transform on the rhythm signal in step 41 to obtain the instantaneous frequency variability of the rhythm signal.

4. A highly sensitive and highly electromagnetic interference resistant brain fatigue detection method as claimed in claim 1, characterized in that: The specific steps for complexity analysis in step 5 are: Step 51: Normalize the complexity of the EEG signal and convert it into the change rate of the cortical EEG signal before and after brain fatigue: Among them, MSE fatigue Represents the multi-scale entropy after fatigue, MSE normal represents the multiscale entropy when not fatigued; Step 52: Take the data change rate of 10s and average it 8 times: Step 53: Based on Rate and Calculate the standard deviation: Where n represents the number of times the average is taken; Step 54: Calculate the standard error SEM of the sample mean based on the standard deviation: Here, N is the number of subjects.

Images