Publicado en línea: 27 feb 2025
Recibido: 14 oct 2024
Aceptado: 12 ene 2025
DOI: https://doi.org/10.2478/amns-2025-0099
Palabras clave
© 2025 Mingmin Kong, published by Sciendo
This work is licensed under the Creative Commons Attribution 4.0 International License.
In recent years, as the accelerated pace of life, the increasingly fierce and cruel social competition, people’s negative psychological emotions are increasing. Many vulnerable people have poor endurance in the face of adverse events, and failing to take timely measures to adjust their emotions will seriously affect their daily life quality. Negative emotions can cause irritability, insomnia, memory loss, inattention and other abnormalities, long-term negative emotions, leading to endocrine disorders, low immunity, physical damage, leading to various physical and mental diseases, and even severe depression [1]. Therefore, the ability to adjust emotions is very important. Studies have shown that the emotional perception of patients with various mental diseases is often deficient. For example, patients with schizophrenia have difficulty perceiving multiple emotional pictures such as sadness and anger [2]. Emotional perception also has a significant positive impact on emotional adjustment strategies [3].
Adrian et al. Research. There are two ways to regulate psychological emotions. One is external adjustment, which is mainly through interpersonal adjustment of talking with people, and through the natural law through outdoor activities. The other is internal regulation, which refers to individuals’ physiological, psychological, and behavioural regulation through psychological cues and physical activities [4,5]. Aerobic exercise is the most commonly used form of emotional regulation. Aerobic exercise regulates emotions and has a cathartic function, releasing stress and making one physically and mentally happy. Aerobic exercise is often used in treatment of depression. Studies have shown that exercise is effective in the treatment of depression, and can significantly improve the exercise in patients with mild depression. In this paper, we analyzed the impact of short and long-termerobic exercise on mood, respectively, and how aerobic exercise improves mood in both immediate and long-term effects.
Studies have shown that during and after aerobic exercise, the secretion of endorphins and dopamine will be accelerated. These two substances will bring happiness and happiness, thus improving mood [7]. The mainstream research is still exploring the influence of exercise on subjects’ emotions through subjective question-and-answer forms such as scales and questionnaires. This form of research explains the influence of exercise on emotion through subjective judgments, which is highly personalized and will significantly interfere with the experimental results. In contrast, this article uses EEG to detect and evaluate the influence of exercise on emotion regulation. This method measures the physiological signals spontaneously generated by the human body, which does not change with subjective thoughts, making the measurement results more objective and reliable. In addition, this paper innovatively uses an emotion recognition and classification method to evaluate the influence of aerobic exercise on emotion regulation. This method has been studied in other people, but has not been studied in the field of sports emotion regulation. The use of emotion recognition allows an intuitive and effective analysis of the subjects’ emotion regulation abilities, and observes the influence of long-term aerobic exercise habits on emotion regulation.
To sum up, this paper will analyze the EEG differences of non-exercise habit group, aerobic exercise group and long-term aerobic exercise group when facing different scenes, and explore the influence of aerobic exercise on emotion regulation ability, so as to provide a new detection method for preventing negative emotions by exercise.
Depression can seriously affect people’s daily life and interpersonal interactions, and even suicidal tendencies can occur in severe cases. At the same time, exercise can regulate mood and improve social functioning and quality of life in depressed patients [12]. Studies show that the suicide rate of depressed patients is about 4.0% to 10.6%, and reports show that more than 46% of those who die by suicide have mental illness [10]. With such a significant impact, depression is difficult to cure, and according to the Chinese Guidelines for the Prevention and Treatment of Depressive Disorders, overall recurrence of depression is as high as 50%-85%, with 50% of patients relapsing within two years of the onset of the illness [11]. Therefore, how to regulate mood and prevent depression is crucial. In this paper, a study on the effect of aerobic exercise on mood regulation was conducted based on the above research background.
Gutmann et al. found that individual Alpha Peak Frequency (iAPF) is a neurophysiological marker of arousal and attentional state. iAPF is expected to return to baseline values 10 to 20 minutes after cessation of exercise. The observed transient increase in iAPF may be a potential neurophysiological marker of improved perception and cognitive function. The observed quick increase in iAPF may be a possible neurophysiological mechanism for improved perceptual and cognitive function. To detect changes in EEG in the alpha band, independent of the effect of heart rate changes on EEG, the experiment in this paper requires that the subject’s heart rate be restored to no more than 10% of its resting state for a maximum of 10 minutes [13]. In addition to time, it is also necessary to capture the association between exercise intensity and emotion, as Panteleimon et al. have summarized 33 studies on the existence of a dose-response relationship between exercise intensity and emotion [14]. These articles used scales to measure sentiment, where for low-intensity exercise, some studies concluded that it has a positive effect on emotion, and some concluded that it does not have a significant impact [15,16]. For moderate-intensity exercise, most of the studies concluded that it effectively reduces negative emotions [17]. While the high-intensity exercise part of the studies finished that it is the most effective in reducing negative emotions, part of the studies concluded that it does not affect. In contrast, another part of the studies concluded that it enhances the production of negative emotions [18,19,20]. The article concludes that the effect of exercise workouts of different intensities may also be related to self-will and physical fitness. The intent of this paper was to investigate the impact of aerobic exercise on mood in most people. Therefore, the acute aerobic training in this paper was chosen as a moderate-intensity aerobic exercise that is acceptable and effective for most people.
Okun’s paper studies the influence of 6 months yoga training on the cognition and quality of life of 65-68 years old healthy elderly. The Mood State Scale was used to assess mood, the Multidimensional Fatigue Scale to detect fatigue, the POMS and State Anxiety Scale to assess depression, and the Health-Related Quality of Life Scale to detect the health-related quality of life. The results showed that yoga does not significantly improve cognitive function, but does significantly improve quality of life and physical fitness [21]. Nathaniel and others. Studied the effect of acute exercise on anger mood, and the experiment took the State-Trait AngerExpression Inventory 2 (STAXI-2), the self-assessment manikin (SAM) and ERP analyses. The results of the scales showed that vigorous exercise could reduce anger and also reduce the induction of anger but does not change the intensity of anger. However, the EEG results showed that exercise had no effect on event-related potentials [22]. The above research mainly analyzed emotions through the subjective consciousness scale detection methods, and this paper mainly studies the influence of exercise on emotions through EEG.
Hall and others. It shows that asymmetry could be used to identify emotional states [23], and in 2010, it was proved that asymmetry of frontal lobe could predict inspiration after exercise [24]. Jennifer et al.’s research. The same asymmetry of forehead Ye Jing EEG is used to characterize the risk of depression [10]. On this basis, Ohmatsu Satoko and others. The combination of 5-HT, treadmill running and EEG signals was analyzed. They showed that treadmill exercise activates serotonin, which led to 1 (7.5-10 Hz) wave asymmetrical increase in frontal lobe and improved negative emotion [9]. Crabbe and others. The asymmetry of EEG signals in forehead and parietal lobe after 30 minutes of moderate intensity pedaling exercise was studied. They proved that acute aerobic exercise usually does not change people’s emotional responses to happy and neutral pictures, so as to reduce their emotional arousal to unpleasant stimuli. By comparison, Sutton and others. It shows that activation in the left frontal lobe reflects the increase in positive emotions, not the increase of the right frontal lobe, which reflects the increase in negative emotions. These studies on asymmetry explain in detail why exercise can regulate emotions, but they don’t solve the problem of whether exercise can regulate emotions uniformly in different environments? In this paper, this problem was studied in aerobic exercise experiments.
At present, most studies on the influence of long-term aerobic exercise on emotion use scales to test the potential of events and changes of signals themselves. For instance, Li Luo and others. Through a large-scale study, it is found that long-term walking and Tai Chi exercise enhance the emotional face recognition of the elderly. Fumoto Masaki and others. The respiratory frequency, energy of ECG signal and EEG signal were compared between the skilled group and the beginner group during the Tai Ji Chuan exercise. They concluded that regular practice of taijiquan exercise could effectively improve physiological functions, especially in terms of improving brain function and using nerves to control muscles. Fumoto Masaki and others. It was found that the content of 5- hydroxytryptamine in subjects increased significantly after exercise. Based on the results of emotion recognition, this paper analyzes the emotional adjustment ability of those who have long-term aerobic exercise habits and those who have no exercise habits. For example, Weser Matthias J et al. The emotion recognition defects of patients with Parkinson’s disease were studied. The study have shown that the patients’ emotional discrimination defects might not be caused by early emotional treatment defects. The event-related potentials of patients and normal subjects were compared under three kinds of performance pictures (positive, negative and neutral). Keane gill and others. Facial Expression Recognition of Patients with Dementia and fvFTD. In this study, an emotion recognition study was conducted, and the patients underwent a series of facial perception tasks and voice emotion recognition tests. The results showed that FFTD patients’ facial expression recognition ability was impaired, but the facial feature recognition function was preserved, and there were emotional recognition obstacles.
These are recent studies on the influence of exercise on emotion and emotion recognition in different groups of people. Previous studies on acute aerobic exercise only discussed whether exercise affects emotion and the reasons for it, but did not discuss whether the mechanisms of exercise’s influence on emotion in different contexts is consistent. For long-term aerobic exercise, previous studies mainly focused on the discussion of scales, ERP analysis, and changes in EEG. However, they have not used emotion recognition to discuss emotion regulation. Therefore, this paper uses emotion recognition to study the influence of exercise on emotion regulation.
In this paper, two experimental protocols were used to compare and analyze the emotional changes of the no-exercise population, the acute exercise population and the long-term aerobic exercise population when faced with different situations.
Subject recruitment. In the acute aerobic exercise experiment, 80 students who did not exercise regularly were recruited as experimental subjects, 40 of whom were the control subjects who did not exercise and 40 of whom were the subjects in the moderate-intensity aerobic exercise group; in the long-term aerobic exercise habit experiment, 80 students were recruited as experimental subjects, 40 of whom were physical education students in the long-term aerobic exercise habit group and 40 of whom were the issues in the long-term aerobic exercise habit group. The experiment on long-term aerobic exercise habit also recruited 80 students as experimental subjects, 40 of whom were physical education students as the experimental group with long-term aerobic exercise habit, and the other 40 were control subjects without exercise habit—all the subjects filled in the physical activity scale to determine whether they had exercise habits or not. Acute aerobic exercise comparison experiment. This experiment illustrated the effect of acute aerobic exercise on mood regulation by comparing the mood changes between the exercise and control groups after the activity. The EEG signals were collected before and after 20 minutes of action (exercise group: moderate-intensity aerobic exercise; control group: sitting and resting) by the International Mood Picture System (IMPS), and the Mood State Model Questionnaire was collected as their subjective mood evaluation for analysis. Exercise intensity was assessed by the Rating of perceived exertion (RPE) and real-time heart rate. Afterwards, the prefrontal asymmetry of the EEG signal was calculated in the frequency domain to compare and analyze the mood changes of the two groups of subjects. The effect of acute aerobic exercise on the subjects’ mood regulation was derived. Long-term aerobic exercise habit comparison experiment. This experiment was conducted by comparing the differences between the subjects in the no-exercise habit group and the long-term aerobic exercise habit group in terms of evoked emotions under the same picture stimuli. The EEG signals of the issues with no exercise habits and the subjects with long-term aerobic exercise habits were collected under positive, negative and neutral picture stimuli. The α and β band signals were decomposed by wavelet transform. The energy asymmetry, root mean square, and recursive energy efficiency of the prefrontal channel was extracted as emotion recognition and classification features, respectively. The best emotion recognition model was constructed based on a backward selection algorithm and ten-fold cross-validation to investigate the effect of long-term aerobic exercise habit on subjects’ emotion regulation ability.
Brain waves are a method of recording brain activity using electrophysiological indicators. When the brain is active, the sum of postsynaptic potentials co-occurring in many neurons forms brain waves. EEG is a general reflection of the electrophysiological activity of brain nerve cells in the cerebral cortex and records the changes in electrical waves during EEG activity. The study by Cacipoop et al. indicates that emotional changes lead to physiological changes in facial muscle activity, brain activity, and autonomic nervous system activity, respectively. Therefore, the experiments in this chapter will reflect individuals’ vigorous exercise by recording the brain’s physiological activity.
Time-domain characteristics
The time domain is the real world, the only part that exists. Although the time-domain features of EEG are not dominant, there still exist many ways to identify the time-series features in different emotional states. Standard time-domain features are as follows: 1) ERP signals. Event-related potentials are a special kind of brain evoked potentials mainly used to analyze emotional changes by inducing potential changes in the brain by specific stimuli. For example, Frantzidis et al. analyzed emotions by the amplitude and latency of emotion-related potentials such as P100, N100, N200, P200, P300, and LPP. 2) Statistical characteristics. Energy, mean, standard deviation, signal difference, normalized difference, etc., are often used to describe the time series of EEG.
Frequency domain features
The frequency domain is used to describe the characteristics of the signal in terms of frequency, and the amount of movement in each given frequency band within a frequency range is displayed in the frequency domain diagram. The most common features in EEG emotion recognition are the energy features in different frequency bands. The most common algorithm used to calculate the discrete Fourier transform is the fast Fourier transform, which can also be computed using the short-time Fourier transform and energy spectral density. The features often extracted from the signal frequency domain are the average energy of each frequency band, the minimum, maximum, and variance. In addition, the average band energy β/α of each channel can also be used for feature extraction.
Time-frequency domain features
The EEG signal is a non-stationary signal, so more information can be obtained by analyzing the dynamic changes of the signal using time-frequency domain methods. The discrete wavelet transform is a common time-frequency domain method for EEG signal analysis, which can obtain more signal features than the Fourier transforms by performing different approximations and detail-level decomposition of the signal according to different frequency ranges while preserving the signal time information. The decomposed signal depends mainly on the wavelet basis, the number of decomposition layers and the sampling rate of the signal, and the time-frequency domain features are extracted based on the decomposed signal. For example, Murugappan et al. used the ‘4'db wavelet function to extract the energy and entropy of the fourth level of detail coefficients (D4) of the EEG corresponding to the α-band, and the subsequent studies also extracted the root mean square of the combination of the β-band and γ-band signals and the energy efficiency of the recursion. The feature extraction in this paper is based on this.
Multichannel fusion features
The features are usually calculated based on the recorded signals of a single electrode, but there are also some features that are calculated by combining the signals of multiple electrodes, which are multichannel fusion features. Common multichannel fusion features are shown below: 1) Asymmetry. EEG asymmetry is usually calculated from the difference in power bands of corresponding electrode pairs, and there also exists a calculation of the difference between features extracted using corresponding electrode pairs, such as kurtosis, maximum value, number of peaks, and other features. 2) MSCE (Magnitude Squared Coherence Estimate ) consists of the power spectral density of different electrode points at The MSCE is calculated from the power spectral density of different electrode points at the same frequency, which is equivalent to the number of correlations in earlier studies.
The raw EEG signals are easily disturbed by the external environment due to their small amplitude of variation, and many non-EEG components are mixed in the raw signals collected, so we need to pre-process the EEG signals before calculating them. In this paper, we investigate the association between aerobic exercise and emotion through EEG signals, so we selected EEG data from prefrontal areas, i.e. channels Fp1, Fp2, F3, F4, F7 and F8. EEG cap localization map is shown in Figure 1.
32-bit EEG cap localization map
The EEG cap selected for the experiments in this paper was a 32-bit EEG cap, set up according to the 10-20 international EEG recording system, the principle of the 10-20 system being that the relative distances between scalp electrodes are expressed in terms of 10% and 20%. The brain is divided into four lobes: the frontal lobe (located in front of the central sulcus. It is responsible for higher functions such as intelligence, emotion, and self-awareness. The language, writing, and motor centres are located in the frontal lobe; the parietal lobe is located behind the central sulcus. It is responsible for bodily sensations (pain, touch, etc.), understanding language and words (visual language centre), interpreting information transmitted from the various senses, and also for spatial perception; the occipital lobe is in the posterior part of the brain, behind the parietal and temporal lobes. The visual cortex is located, it mainly processes information related to vision, such as colour and luminance; the temporal lobe is located below the lateral interstitial sulcus. It is responsible for hearing, understanding language (auditory language centre), memory (hippocampus), etc.
During the acquisition of EEG signals, the signal that scalp electrodes can measure is the potential difference, i.e., the difference between the acting electrode and the reference electrode. Therefore, to obtain the ideal raw data, the least active electrode point will be used as the reference in this paper. In this paper, the bilateral mastoid averaging reference technique is the most commonly used reference tool. Bilateral mastoid averaging is a method of averaging the reference electrodes on the mastoid behind the ear on both sides and then averaging the two electrodes as a reference signal, effectively avoiding distribution distortion. The principle is as follows.
There is a ground electrode at a location on the top of the head, and its voltage value is set to X. The original EEG signal amplitude at electrode point A is set to Y. Then the recorded EEG amplitude is the voltage difference between the recording electrode and the ground electrode α=Y-X; the original signal amplitude of the left and right mastoids is set to and L R, respectively, and the recorded signal amplitude l=L-X and r=R-X. After conversion to the average reference of bilateral mastoids, the signal amplitude at position A becomes
The reference electrodes used in this paper are A1 and A2 channels, which are the average reference method for bilateral mastoids.
Artifact detection is of great significance in the process of EEG acquisition, which can determine the reliability and stability of EEG signals. Artifacts need to be detected in the data after EEG signal acquisition. Because the frequency of EEG signal is relatively low, high-frequency signals are fused in the EEG signal, resulting in distortion of the EEG signal. After the EEG signal is collected, the original information needs to be filtered, and the filtering method generally adopted is as follows:
Adaptive filtering: By analyzing the characteristics of the input signal, the filtering parameters can be adjusted according to the environmental characteristic factors to achieve effective filtering of the signal. The real-time change of the input signal is automatically adjusted to realize the minimum difference between the adjusted signal and the expected signal. The EEG signal is used as the input, the noise signal of the artifact is used as the reference, and the denoised EEG signal is realized by the method of the difference of the two. The signal denoising process is as follows:
Where x(n) represents the denoised signal, S(n) is the original EEG signal, and m(n) is the artifact signal, which needs to be filtered and removed. According to the adaptive denoising characteristics, the output signal y(n) needs to be adaptively adjusted by x(n), and the formula is as follows:
Wavelet denoising: Wavelet denoising is a method that uses Fourier transform to filter non-stationary signals, which can capture the instantaneous characteristics of brain signals. The model of wavelet denoising is expressed as follows:
Wavelet denoising is to decompose the signal into wavelet basis functions with different scales, and then convolute these wavelet basis functions with the original signal to obtain the components of the original signal at different times and frequencies, thus revealing the characteristics of the signal at different time scales and frequency scales. The denoising performance is determined by the signal-to-noise ratio and the fitting coefficient, and the signal-to-noise ratio formula is as follows:
The fitting coefficient represents the retention of the signal portion of the actual value of the EEG signal, and is expressed by the following formula.
Empirical mode decomposition: In essence, empirical mode decomposition is also an adaptive denoising method. Compared with the traditional adaptive denoising method, it does not need external functions and can ensure high stability in signal decomposition. The original signal is decomposed into a variety of local frequency components, and the local characteristics of the signal can be effectively obtained. The signal decomposition is shown in the following formula:
By fitting the upper and lower envelopes of X (t), the upper envelope is fitted to the maximum value by
Subtract m1(t) from x(t) to obtain the remaining h1(t).
The EMD processing flow is shown in Figure 2:
Empirical Mode Decomposition Process
The EEG consists mainly of various rhythmic electrical activities. According to frequency, EEG waves are generally classified into the following five categories.
δ wave: the frequency is 0.5~3 Hz and the amplitude is 10~20 V. It is mainly located between the frontal and parietal lobes of the brain. It is the basic waveform of the infant brain and can also be seen in physiological slow-wave sleep and pathological coma states. θ wave: the frequency is 4~7 Hz and the amplitude is 20~40 V. It is mainly located in the temporal and parietal lobes. This waveform indicates that the brain is in a thinking state, and is the basic waveform for preschool children and can also be seen in adults during the drowsy state. α wave: With a frequency of 8-13 Hz and an amplitude of 10-100 V, it has a wide distribution and can be detected in all locations of the head, especially in the occipital region. This wave is commonly found in adult EEG signals and has a significant rhythm compared to other waves. β wave: the frequency is 14~30 Hz and the amplitude is below 20 V. It is mainly located in the frontal area. It is easily generated when a person is awake or when he or she is nervous or excited, so to a certain extent it can also reflect the attention of the brain and the emotional state of a person. γ wave: It is the highest frequency component of the EEG signal, mainly located in the frontal and central regions. It is generated when a person focuses on something or is alert, and is related to the higher task processing and cognitive processing activities of the human brain.
The experimental content of this paper was completed by young university students in the awake state, and pictures were used to induce emotion and motor regulation during the experiment, so α and β waves were selected for calculation.
The KNN algorithm is based on the principle of finding the closest K training samples in the training set based on the distance measure and then predicting the target based on the information of these K most immediate values. KNN includes three elements: K, distance and classification rules. Distance measures the physical distance between two points as follows:
KNN describes the distance of two kinds of x1 and x2, and the classification result can be applied well.
The SVM algorithm is a Gaussian kernel function, also known as a radial basis function, which essentially maps the sample points to an infinite-dimensional feature space, making the linearly indivisible data linearly divisible is very practical for EEG. The principle of the transformed information is shown in Equation (11).
Where
The effect of acute aerobic exercise on negative emotional stimuli was illustrated by an experimental exploration of the dynamic changes in the exercise group and the control group before and after aerobic exercise. In this experiment, the index used to compare emotional changes in this chapter is the asymmetry of prefrontal lobe. Based on this study, the following two aspects were analyzed: (1) To compare the change in forehead asymmetry between the two groups before and after 20 minutes of moderate-intensity aerobic exercise, to analyze whether acute aerobic exercise has an improved effect on the subjects’ mood. (2) compare the energy changes of left and right brain before and after exercise in negative and neutral conditions, and analyze whether acute aerobic exercise can also improve the mood.
In this experiment, 80 college students with normal vision (or corrected vision) and body mass index in the normal range of the IPAQ scale were recruited to participate in the experiment. Due to the high reliability and validity of IPAQ scale, the retest reliability was 0.82, and the subjects were used to screen the subjects. The IPAQ was used to filter the issues because of its high reliability and validity. In this experiment, 80 subjects without exercise habits were selected, and 66 subjects were selected in the later data processing, because the EEG signals did not meet the adoption standards during the experiment. These subjects were divided into exercise groups: 30 (15 males and 15 females) and a control group: 36 (19 males and 17 females). Table 1 shows the details of the subjects.
Pleasantness and arousal rate of stimulus pictures before and after exercise
| Mean value (pleasantnes) | Standard deviation (pleasantness) | Mean value (arousal rate) | Standard deviation(arousal rate) | ||
|---|---|---|---|---|---|
| Pre-test experiments | Negative Picture | 2.88 | 0.97 | 5.80 | 1.51 |
| Neutral Picture | 4.99 | 0.04 | 3.68 | 1.11 | |
| Post-test experiments | Negative Picture | 2.87 | 0.96 | 5.76 | 0.91 |
| Neutral Picture | 4.99 | 0.04 | 3.26 | 1.17 |
Heart rate is one of the most commonly used and easily measured physiological indicators in exercise physiology. The degree of variation in heart rate reflects the intensity of exercise; therefore, heart rate was used to measure exercise intensity in this experiment, according to the conclusion of Tanaka. H, the maximum heart rate calculation is closely related to age, so the maximum heart rate was estimated using equation (3). The Q value was calculated according to equation (4), and when Q < 50% indicates low intensity exercise, 50% < Q < 75% indicates moderate-intensity exercise, and Q > 75% is high intensity exercise. During this experiment, an exercise bracelet was used to observe the real-time heart rate. The exercise timer was started when the subject’s real-time heart rate exceeded 50% of the maximum heart rate. The subject was asked and filled in the RPE every 2 minutes during the exercise process to determine the subjectively perceived exercise intensity during the exercise.
The subjective effect was assessed using the PAD-P questionnaire, a psychological model developed by Albert Mehrabian and James A. Russell to describe and measure emotional states. The PAD uses three numerical dimensions, pleasure, arousal, and dominance, to represent all emotions. The PAD-P scale measures how much a person feels something. PAD-A measures a person’s energy or hypnotic state. PAD-D (Dominance-submissiveness ) is expressed as control and dominance-controlling or submissive. This study was more concerned with emotional pleasantness and therefore used the Pleasantness Emotion Model Scale.
The overall flow of this experiment is shown in the Figure 3. The experimental process of the two groups of subjects in the session other than exercise was kept the same, which could effectively avoid the influence of other factors on the experimental results. The emotional stimulation experiment was based on the picture stimulation to induce emotion, and the picture was played for 6 seconds to fully induce and record the subjects’ emotional responses.
Overall experimental flow chart
In Table 2, the subjects’ personal information, RPE and PAD-P scores were recorded. According to Table 2, we can find that the age of the subjects in the exercise group and the control group are basically the same, the difference in BMI is small, and the physical quality of the subjects in the two groups is basically the same. Secondly, the RPE scores of all subjects in the exercise group showed that the subjectively perceived exercise intensity was 5.57, which was in line with the requirement of moderate intensity.
Subjects’ basic personal information, RPE and PAD-P scale scores
| Average value (Sports group, 30 people) | Standard deviation (Sports group, 30 people) | Average value (Control group, 36 people) | Standard deviation(Control group, 36 people) | |
|---|---|---|---|---|
| Age | 20.22 | 1.19 | 19.56 | 1.32 |
| Height | 167.38 | 8.37 | 169.33 | 8.28 |
| Body weight | 56.78 | 8.09 | 60.47 | 9.92 |
| BMI | 20.08 | 1.71 | 21.02 | 2.56 |
| PRE | 5.57 | 0.77 | / | / |
In addition, the asymmetry of EEG was calculated. Firstly, the power spectral density of the original signal is obtained by short-time Fourier transform, and the total energy of the signal is obtained according to the power spectral density. After removing the baseline energy, the prefrontal EEG signal asymmetry Asym is the difference between the energy of the F7 channel and the F8 channel EEG, which is shown in Equation (14).
Figure 4 shows the change in forehead EEG asymmetry in the two groups of subjects. The problem of forehead asymmetry in exercise group. The forehead asymmetry was significantly increased in issues in the motor group under negative stimuli and slightly increased under neutral stimuli. In both cases, the asymmetric increase of exercise group indicates that acute aerobic exercise can improve mood in both cases. Compared with the magnitude of improvement in mood in both contexts, it was found that acute aerobic exercise effectively suppressed negative mood. The forehead of the controlled object is asymmetrical. Under the negative and neutral stimulation, the asymmetry of the forehead of the control subjects decreased slightly, which indicated that the mood of the control subjects did not improve in either case after meditation. The rise in forehead asymmetry represents an increase in a positive mood, suggesting that acute aerobic exercise can effectively regulate mood.
EEG signal asymmetry in two groups of subjects
By calculating the energy of the F7 and F8 channel EEG, the energy changes in the left prefrontal and right prefrontal areas of the subjects in the motor group under negative and neutral stimulation can be seen in Figure 5. (a) Under negative stimulation, the energy in the left prefrontal area was decreased and the energy in the right prefrontal area was increased. Relative to the pre-test experiment, activation in the left prefrontal area was enhanced in the post-test experiment, producing positive emotions, while exercise inhibited activation in the right prefrontal area, reducing the production of negative emotions. (b) With neutral stimulation, there was a decrease in energy in both prefrontal areas and enhanced activation in both left and right brain areas, but the overall presentation was positive emotion, with exercise contributing to more positive emotion production.
Signal energy of the F7 and F8 channels of subjects in the exercise group
Under the stimulation of the same picture, by comparing the emotion recognition rates between the long-term aerobic exercise habit group and the non-exercise habit group, the similarities and differences of emotion regulation ability were analyzed. Emotion recognition in EEG has been studied in many ways. Nevertheless, this paper makes a more comprehensive analysis of EEG frequency band, EEG signal channels, classification algorithms and optimal feature subsets. In this experiment, both groups of subjects received the same picture emotional stimulation. Two classification algorithms (SVM algorithm and KNN algorithm) were used to extract three types of features (energy asymmetry, root mean square and recursive energy efficiency) for emotion recognition of EEG signals. Finally, we analyzed and compared the emotion recognition rates of the two groups in each aspect of the optimal emotion recognition model to illustrate the effect of long-term aerobic exercise habits on emotion regulation ability.
According to Table 3, there is a slight difference in the accuracy of the EEG of the long-term aerobic exercise habit group after using the two classification algorithms for emotion recognition. For the positive EEG data, the correct average rate of the SVM algorithm in the €-bandnd í - band was slightly higher than that of the KNN algorithm. In contrast, for the negative EEG data, the correct average rate of the KNN algorithm in the € and ę-bands was slightly higher than that of SVM algorithm. However, for the AUC scores, the SVM algorithm was higher than the KNN algorithm for all classifications. There is no significant difference in accuracy between the two algorithms. Still, the AUC scores of the SVM algorithm are significantly higher than those of the KNN algorithm, so it can be concluded that the SVM algorithm is superior to the KNN algorithm in recognition of emotions in the EEG of the long-term aerobic exercise habit group. KNN and SVM models are used to compare the performance of Positive and Negative with Precision, Recall and F1 values respectively. The performance comparison results are shown in Figures 6-8.
Number of features, confusion matrix, accuracy and AUC of the optimal subset of the long-term aerobic exercise habit group
| Number | FP | TP | FN | TN | ACC(%) | AUC | |||
|---|---|---|---|---|---|---|---|---|---|
| Positive | β | SVM | 12 | 53 | 112 | 49 | 115 | 69.00 | 75.56 |
| KNN | 13 | 46 | 118 | 57 | 107 | 68.60 | 68.54 | ||
| α | SVM | 11 | 61 | 102 | 55 | 108 | 64.42 | 67.96 | |
| KNN | 14 | 53 | 110 | 61 | 102 | 65.03 | 64.88 | ||
| α-β | SVM | 18 | 54 | 109 | 32 | 130 | 73.54 | 79.92 | |
| KNN | 16 | 36 | 126 | 51 | 111 | 73.15 | 73.15 | ||
| Negative | β | SVM | 12 | 61 | 105 | 47 | 118 | 67.37 | 71.10 |
| KNN | 14 | 50 | 116 | 59 | 107 | 67.17 | 67.11 | ||
| α | SVM | 13 | 65 | 101 | 51 | 115 | 65.06 | 68.52 | |
| KNN | 11 | 50 | 116 | 62 | 104 | 66.27 | 66.25 | ||
| α-β | SVM | 16 | 58 | 106 | 33 | 131 | 73.26 | 78.75 | |
| KNN | 16 | 38 | 126 | 50 | 115 | 74.25 | 73.11 |
Comparison of Precision rates of KNN and SVM
Comparison of Recall rates of KNN and SVM
Comparison of F1 rates of KNN and SVM
In this paper, the results of two experiments, acute aerobic exercise and long-term aerobic exercise habit, confirmed that aerobic exercise has a significant positive effect on psycho-emotional regulation. In the acute aerobic exercise experiment, the prefrontal asymmetry of subjects in the exercise group increased significantly after acute aerobic exercise relative to the pre-test experiment, while the subjects in the control group did not increase significantly, indicating that acute aerobic exercise has the effect of regulating the subjects’ mental emotion; while in the long-term aerobic exercise experiment, the mental emotion recognition rate of subjects in the long-term aerobic exercise habit group were higher than those in the no-exercise habit group, and the emotional Therefore, it is believed that long-term aerobic exercise is beneficial to increase the mental-emotional regulation ability. However, due to the limitations of today’s equipment the noise generated during the activity was too loud to accurately capture the effective EEG signal, so the immediate changes of EEG during exercise could not be analyzed for the time being. These factors have not been studied due to the limitations of time and other aspects, and it is hoped that subsequent work will have some refinement of them.