Instant teaching quality evaluation method based on brain-computer fusion, storage medium and electronic terminal

Abstract

The invention relates to the technical field of intelligent education, and particularly discloses an instant teaching quality evaluation method based on brain-computer fusion, a storage medium and an electronic terminal.The method comprises the steps that original electroencephalogram signals of 14 designated positions of a tested student are collected through an international standard 10-20 lead system and preprocessed; and a filter based on a Butterworth transfer function is designed to carry out zero-phase band-pass filtering on the frequency band of 4-45Hz, and electroencephalogram signal artifacts are removed. And extracting difference entropies from the preprocessed signals according to frequency bands, inputting the difference entropies as training features into a neural network model, providing a teaching quality evaluation classification model based on wake-up degree and titer degree by optimizing parameter training, and outputting a quantitative evaluation result. By collecting and analyzing the electroencephalogram signals of the students in the teaching process, the teaching quality is objectively fed back in real time, the problems that traditional teaching evaluation is high in subjectivity and poor in precision are effectively solved, and teachers are helped to improve the teaching level. Meanwhile, the invention provides an electronic terminal capable of executing the method and a computer program storage medium.

Description

Technical Field

[0001] The present invention belongs to the field of smart education technology, and specifically relates to an instant teaching quality assessment method based on brain-computer fusion, a storage medium and an electronic terminal. Background Art

[0002] In the current field of primary and secondary education, teaching quality assessment is an important means to improve the effectiveness of teacher education and optimize teaching strategies. There are various traditional teaching quality assessment methods, including teaching observation, questionnaire survey, class observation and evaluation, student evaluation, educational experience summary, and quantitative and qualitative evaluation. These methods comprehensively evaluate teaching quality through direct observation, student feedback, teacher mutual evaluation and experience summary. Paper-and-pencil tests, classroom observation, and homework correction focus on the evaluation of knowledge mastery and learning process, while project-based evaluation, diversified evaluation and reflective evaluation pay more attention to the improvement of students' comprehensive ability and self-cognition. Evaluation based on learning analysis uses big data technology to deeply mine student learning data and provide support for precise teaching. However, these traditional teaching quality assessment methods often rely on students' test scores, teachers' self-evaluation or peer evaluation, and have defects such as strong subjectivity, long feedback cycle, and inability to adjust teaching strategies in real time. Especially in the classroom teaching process, it is difficult for teachers to obtain students' understanding and reaction to the teaching content in real time, making it difficult to make targeted adjustments.

[0003] In recent years, with the rapid development of brain-computer interface technology, a new direction has been proposed to evaluate the teaching quality by collecting and analyzing students' EEG signals. As a direct reflection of brain activity, EEG signals can reveal students' cognitive processing of teaching content in real time and objectively. Brain-computer fusion is an inevitable trend in the development of brain-computer interface technology. In the brain-computer fusion system, the brain and the brain, and the brain and the machine are not only brain-computer interconnected at the signal level, but also need to realize the fusion of the brain's cognitive ability and the machine's computing ability. This fusion system has three significant characteristics: a more comprehensive perception of the organism, including the understanding of apparent behavior and the decoding of neural signals; the organism is also the perception, calculation and execution body of the system, and the information interaction channel with other parts of the system is bidirectional; multi-level and multi-granular comprehensive utilization of the capabilities of the organism and the machine to achieve a great enhancement of system intelligence. In the field of primary and secondary education, brain-computer fusion technology has demonstrated its unique application value. By directly connecting human brain activity to computers or other external devices, the technology can monitor students' learning status and concentration in real time, thereby helping teachers adjust teaching content and rhythm according to students' learning effects and acceptable challenges, and realize personalized teaching. Wang et al. examined the influence of learning-related factors on attention measured by portable electroencephalogram and developed an attention classification algorithm; Martinez et al. proposed a low-cost platform for assessing students' attention and estimating their learning performance; Fan et al. analyzed students' emotions towards teachers based on support vector machines. Based on brain-computer fusion technology, teachers can adjust the difficulty of learning materials according to students' learning feedback and provide real-time guidance to better meet students' learning needs.

[0004] However, there are still some challenges in the application of existing technologies in teaching quality evaluation, such as high equipment complexity, difficulty in signal acquisition and processing, inaccurate evaluation model, and difficulty in extracting key features. In order to fuse multi-scale features and extract contextual information, reduce the time and space complexity of the model, and achieve more accurate teaching quality evaluation, the present invention proposes a method for real-time teaching quality evaluation based on brain-computer fusion. It is not based on a specific fixed public data set, but can predict the current student learning status based on the EEG signals of the students at any time, and comprehensively evaluate the current teaching quality of the teacher from the perspective of valence and arousal. The framework gradually changes from an initial shape to a three-dimensional grid model, which is more refined and realistic, and achieves good classification effect. Summary of the invention

[0005] The purpose of the present invention is to provide a real-time teaching quality assessment method, storage medium and electronic terminal based on brain-computer fusion, so as to realize accurate and reliable assessment of the current teacher's teaching quality by non-invasively collecting the EEG signals of the students. The present invention mainly solves the following technical problems: first, baseline drift removal, bandpass filtering and bad track interpolation are performed on the self-collected noisy original EEG signals; second, a teaching quality assessment model based on frequency band differential entropy is designed, and a quantitative calculation method is proposed.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for evaluating the quality of real-time teaching based on brain-computer fusion, comprising the following steps:

[0008] S1: Based on the international 10-20 lead standard, the consumer-grade EEG acquisition device Emotiv Epoc X dry electrode cap was deployed to collect 14 channels of raw EEG signals from the surface of the brain of the students tested through the dry electrode device, and the data was stored in the form of a two-dimensional array by channel;

[0009] S2: pre-processing the collected raw EEG signals, removing baseline drift, bandpass filtering and artifacts, and interpolating the detected bad tracks to generate initial processable EEG signals;

[0010] S3: Extract the differential entropy of the preprocessed EEG signal in frequency bands as the local key signal feature; and stack it into a three-dimensional form as a data set according to the set rules;

[0011] S4: Input the data set into the pre-built neural network model, adjust the neural network model parameters until the accuracy threshold requirement is met, and propose a teaching quality evaluation model to output the evaluation score results.

[0012] Furthermore, the collection of original EEG signals of the students under test described in step S1 is based on the international 10-20 lead system standard, and the frontal midpoint (Fz), central point (Cz), vertex (Pz), occipital point (Oz), left frontal pole (Fp1), right frontal pole (Fp2), left frontal (F3), right frontal (F4), left central (C3), right central (C4), left temporal middle (T3), right temporal middle (T4), left occipital (O1), and right occipital (O2) are selected as a 14-channel dry electrode device, and two reference sensors are connected to the earlobes in a wet electrode manner to collect the original EEG signals of the students under test; the Bluetooth frequency for collecting the original EEG signals is 2.4 GHz, and the sampling frequency is set to 128 Hz.

[0013] Furthermore, the baseline drift removal in step S2 is a wavelet transform that decomposes and reconstructs the original EEG signal according to rules to remove redundant signals; the wavelet transform is implemented by the following formula:

[0014]

[0015] Where WT(a,τ) represents the signal after wavelet transformation, s(t) is the initial EEG signal, Represents the wavelet basis function; the variables in the wavelet transform include displacement τ and scale a. Scale a determines the analysis accuracy of the wavelet transform in the frequency domain, and displacement τ determines the position of the wavelet function in the time domain. The wavelet basis function selects Daubechies wavelet to improve the frequency domain localization capability.

[0016] Furthermore, the filtering described in step S2 uses a zero-phase bandpass filter based on a Butterworth transfer function to filter out a 4 Hz-45 Hz signal. The Butterworth transfer function H(s) is implemented by the following formula:

[0017]

[0018] Where ω 0 is the center frequency of the filter (set to 25Hz), Q is the filter quality factor, ω s is the angular frequency containing complex variables, and K is the actual gain of the filter; the EEG signal of the processing type is a non-stationary digital signal, which is composed of a series of second-order filters to improve the performance and stability of the filter; the frequency response curve of the Butterworth filter in the passband is very flat and has no ripples, which means that the amplitude and phase changes of the signal in the passband are very small, thereby ensuring the fidelity of the signal; at the same time, it provides a very high suppression ratio, effectively filtering out unnecessary frequency components such as noise and clutter, which is very beneficial to improving the signal-to-noise ratio of the EEG signal;

[0019] The artifact removal described in step S2 is based on ICA. ICA is a blind signal separation technology that can separate the source signal from the mixed signal by finding signal components that are independent in time and fixed in space. According to the time domain waveform, frequency domain characteristics, scalp topography and other information of the independent components, the artifact components are identified. The artifact components usually have specific waveform characteristics (such as spikes, periodic fluctuations, etc.) and scalp distribution characteristics (such as concentrated near the eyes, muscles, etc.). The processed artifact components are set to zero or appropriately weighted, and then the EEG signal is reconstructed to obtain the EEG signal after the artifacts are removed.

[0020] Furthermore, the steps of extracting the differential entropy of the pre-processed EEG signal by frequency bands in step S3 are as follows:

[0021] Specifically, the frequency band feature extraction is achieved through the Butterworth bandpass filter designed in the previous article. The EEG signal is decomposed into four frequency bands: theta (4-8Hz), alpha (8-14Hz), beta (14-31Hz) and gamma (31-45Hz). For each channel of each trial, the first 3 seconds are extracted as the baseline signal and the remaining part is extracted as the trial signal. For each frequency band signal (theta, alpha, beta, gamma) after filtering, its differential entropy is calculated. Differential entropy is a measure of signal complexity, which is calculated by calculating the mean square error of the signal and using the following formula:

[0022]

[0023] Where e is the natural exponent, σ is the standard deviation of the signal sequence; for the reference signal, the differential entropy of each frequency band is obtained by dividing the signal into three segments (128 sample points in each segment), calculating the differential entropy separately and then taking the average value. For the test signal, the differential entropy of each frequency band is calculated once per second (i.e., every 128 sample points), so for each frequency band, a differential entropy value is obtained per second, for a total of 60 seconds, so each frequency band has 60 differential entropy values. The calculated two-dimensional data of the differential entropy of the frequency bands are stacked along the new dimension to form a three-dimensional data set for subsequent classification processing.

[0024] Furthermore, the convolutional neural network model described in step S4 is used to process 4-band EEG signal data and classify it, and the model includes in sequence: an input layer for receiving data; a first two-dimensional convolution block, configured with 4 input channels, 24 output channels, 5*5 convolution kernels and ReLU activation functions; a second two-dimensional convolution block, configured with 24 input channels, 128 output channels, 2*2 convolution kernels, ReLU activation functions and 2*2 maximum pooling layers; a first fully connected layer block, with connection dimensions of 128*2*2 input features, output dimensions of 1*96 features and ReLU activation functions; a second fully connected layer block, with connection dimensions of 1*96 input features and output dimensions of 1*2 features; a Dropout layer with a dropout rate of 0.5 is performed after the second fully connected block to form a forward propagation path of the model to achieve classification processing of EEG signal data. The loss function in the neural network model is set to cross entropy, the Adam optimization strategy is selected as the optimizer, the learning rate in the optimizer is set to 0.00001, the time series length is set to 8, and the batch_size is set to 32; finally, the three-dimensional test EEG signal data set composed of differential entropy is sent to the training to obtain the neural network model, and the classification test is performed on the two indicators of Valence and Arousal, and the accuracy threshold is set to 0.8;

[0025] In order to quantify the teaching quality evaluation results, a teaching quality evaluation method based on Valence and Arousal is proposed. The main formula is as follows:

[0026] Score=0.4*Valence+0.6*Arousal

[0027] Among them, the quantization range of Valence is 0-9, and the quantization range of Arousal is 0-9.

[0028] After reaching the accuracy threshold, the teaching quality evaluation score is calculated for the EEG signal of the current student: the scoring quantitative index range is 0-9, where 0-3.0 is fair, 3.0-6.0 is good, and 6.0-9.0 is excellent.

[0029] In a second aspect, the present invention provides an electronic terminal, comprising a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps of any one of the methods described in the first aspect.

[0030] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any one of the methods described in the first aspect.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] (1) The present invention innovatively applies brain-computer interface technology to the field of teaching quality assessment, which can effectively improve the real-time and accuracy of the assessment. By comprehensively adopting the real-time EEG signals of the test students through multiple channels, it can provide more detailed learning EEG data. It has strong anti-interference ability for baseline drift and motion artifacts that often appear in EEG signals collected by wearable devices; compared with the existing classification model based on machine learning, it selects key features and reorganizes the data set according to the set rules, which is conducive to further mining the key information in the signal and generating a more robust teaching quality assessment model.

[0033] (2) The neural network model provided by the present invention cancels the commonly used rhythm ratio as the basis for evaluation, and instead proposes a teaching quality evaluation method based on valence and arousal, thereby improving the accuracy and comprehensiveness of data quantification indicators. Therefore, it has the ability to monitor students' learning status, helping teachers to deeply understand students' learning needs and cognitive status, thereby adjusting teaching content, methods and rhythm, and realizing personalized teaching plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 is a flow chart of a method for evaluating instant teaching quality based on brain-computer fusion provided by an embodiment of the present invention;

[0035] Figure 2 This is a 14-channel original EEG signal data diagram based on the international 10-20 lead standard provided by an embodiment of the present invention;

[0036] Figure 3 This is a diagram showing the effect of removing the baseline drift after the original EEG signal containing baseline drift is subjected to wavelet transformation provided by an embodiment of the present invention;

[0037] Figure 4 is a schematic diagram of an attention assessment network provided by an embodiment of the present invention;

[0038] Figure 5 It is a teaching quality evaluation score curve diagram in one trial phase provided by an embodiment of the present invention; wherein the X-axis is the sampling point and the Y-axis is the teaching quality quantitative evaluation score value.

[0039] in, Figure 2 Numbers 1-14 are channel names, and the acquisition channels are: midfrontal point (Fz), central point (Cz), vertex (Pz), occipital point (Oz), left frontal pole (Fp1), right frontal pole (Fp2), left frontal (F3), right frontal (F4), left central (C3), right central (C4), left middle temporal (T3), right middle temporal (T4), left occipital (O1), and right occipital (O2). DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] Example:

[0042] See also Figures 1 to 5 As shown,

[0043] A method for evaluating the quality of instant teaching based on brain-computer fusion, comprising the following steps:

[0044] Step 1: Deploy the consumer-grade EEG acquisition device Emotiv Epoc X dry electrode cap based on the international 10-20 lead standard, collect the original EEG signals of 14 channels on the surface of the brain of the test students through the dry electrode device, and store the data in the form of a two-dimensional array by channel;

[0045] The original EEG signals of the students were collected based on the international 10-20 lead system standard. The frontal midpoint (Fz), central point (Cz), vertex (Pz), occipital point (Oz), left frontal pole (Fp1), right frontal pole (Fp2), left frontal (F3), right frontal (F4), left central (C3), right central (C4), left temporal middle (T3), right temporal middle (T4), left occipital (O1), and right occipital (O2) were selected as the 14-channel dry electrode device. Two reference sensors were connected to the earlobes in wet electrode mode to collect the original EEG signals of the students. The Bluetooth frequency for collecting the original EEG signals was 2.4 GHz, and the sampling frequency was set to 128 Hz.

[0046] Step 2: Preprocess the collected raw EEG signals by removing baseline drift, bandpass filtering and artifacts, and interpolate the detected bad tracks to generate initial processable EEG signals;

[0047] The baseline drift is a wavelet transform that decomposes and reconstructs the original EEG signal according to the rules to remove redundant signals. The wavelet transform is implemented by the following formula:

[0048]

[0049] Where WT(a,τ) represents the signal after wavelet transformation, s(t) is the initial EEG signal, Represents the wavelet basis function; the variables in the wavelet transform include displacement τ and scale a. Scale a determines the analysis accuracy of the wavelet transform in the frequency domain, and displacement τ determines the position of the wavelet function in the time domain. The wavelet basis function selects Daubechies wavelet to improve the frequency domain localization capability;

[0050] The filtering uses a zero-phase bandpass filter based on the Butterworth transfer function to filter out the 4Hz-45Hz signal. The Butterworth transfer function H(s) is implemented by the following formula:

[0051]

[0052] Where ω 0 is the center frequency of the filter (set to 25Hz), Q is the filter quality factor, ω s is the angular frequency containing complex variables, and K is the actual gain of the filter; the EEG signal processing type is a non-stationary digital signal composed of a series of second-order filters to improve the performance and stability of the filter; the frequency response curve of the Butterworth filter in the passband is very flat and has no ripples, which means that the amplitude and phase changes of the signal in the passband are very small, thereby ensuring the fidelity of the signal; at the same time, it provides a very high suppression ratio and effectively filters out unnecessary frequency components, such as noise and clutter, which is very beneficial for improving the signal-to-noise ratio of EEG signals.

[0053] Artifact removal is based on ICA. ICA is a blind signal separation technology that can separate the source signal from the mixed signal by looking for signal components that are independent in time and fixed in space. According to the time domain waveform, frequency domain characteristics, scalp topography and other information of the independent components, the artifact components are identified. Artifact components usually have specific waveform characteristics (such as spikes, periodic fluctuations, etc.) and scalp distribution characteristics (such as concentrated near the eyes and muscles, etc.). The processed artifact components are set to zero or appropriately weighted, and then the EEG signal is reconstructed to obtain the EEG signal after artifact removal.

[0054] Step 3: Extract the differential entropy of the preprocessed EEG signal by frequency band as the local key signal feature; and stack it into a three-dimensional form as a data set according to the set rules;

[0055] The steps of extracting the differential entropy of the EEG signal after preprocessing by frequency band are as follows:

[0056] Specifically, the frequency band feature extraction is achieved through the Butterworth bandpass filter designed in the previous article. The EEG signal is decomposed into four frequency bands: theta (4-8Hz), alpha (8-14Hz), beta (14-31Hz) and gamma (31-45Hz). For each channel of each trial, the first 3 seconds are extracted as the baseline signal and the remaining part is extracted as the trial signal. For each frequency band signal (theta, alpha, beta, gamma) after filtering, its differential entropy is calculated. Differential entropy is a measure of signal complexity, which is calculated by calculating the mean square error of the signal and using the following formula:

[0057]

[0058] Where e is the natural exponent, σ is the standard deviation of the signal sequence; for the reference signal, the differential entropy of each frequency band is obtained by dividing the signal into three segments (128 sample points in each segment), calculating the differential entropy separately and then taking the average value. For the test signal, the differential entropy of each frequency band is calculated once per second (i.e., every 128 sample points), so for each frequency band, a differential entropy value is obtained per second, for a total of 60 seconds, so each frequency band has 60 differential entropy values. The calculated two-dimensional data of the differential entropy of the frequency bands are stacked along the new dimension to form a three-dimensional data set for subsequent classification processing.

[0059] Step 4: Input the data set into the pre-built neural network model, adjust the neural network model parameters until the accuracy threshold requirement is met, and propose a teaching quality evaluation model to output the evaluation score results.

[0060] The neural network model is used to process and classify 4-band EEG signal data. The model includes: an input layer for receiving data; a first two-dimensional convolution block, configured with 4 input channels, 24 output channels, 5*5 convolution kernels and ReLU activation functions; a second two-dimensional convolution block, configured with 24 input channels, 128 output channels, 2*2 convolution kernels, ReLU activation functions and 2*2 maximum pooling layers; a first fully connected layer block, with connection dimensions of 128*2*2 input features, output dimensions of 1*96 features and ReLU activation functions; a second fully connected layer block, with connection dimensions of 1*96 input features and output dimensions of 1*2 features; a Dropout layer with a dropout rate of 0.5 is performed after the second fully connected block to form a forward propagation path of the model to achieve classification processing of EEG signal data. The loss function in the neural network model is set to cross entropy, the Adam optimization strategy is selected as the optimizer, the learning rate in the optimizer is set to 0.00001, the time series length is set to 8, and the batch_size is set to 32; finally, the three-dimensional test EEG signal data set composed of differential entropy is sent to the training to obtain the neural network model, and the classification test is performed on the two indicators of Valence and Arousal, and the accuracy threshold is set to 0.8;

[0061] In order to quantify the teaching quality evaluation results, a teaching quality evaluation method based on Valence and Arousal is proposed. The main formula is as follows:

[0062] Score=0.4*Valence+0.6*Arousal

[0063] Among them, the quantization range of Valence is 0-9, and the quantization range of Arousal is 0-9.

[0064] After reaching the accuracy threshold, the teaching quality evaluation score is given for the EEG signal of the current student under test: the scoring quantitative index range is 0-9, where 0-3.0 is fair, 3.0-6.0 is good, and 6.0-9.0 is excellent.

[0065] The electronic terminal proposed in the present invention includes a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps of any method described in the first aspect; and a readable storage medium, on which a computer program is stored, which implements the steps of any method described in the first aspect when executed by the processor.

[0066] It will be appreciated that in the development of any actual implementation, as in any engineering or design project, numerous implementation-specific decisions may be made. Such a development effort may be complex and time-consuming, but will be a routine task of design, fabrication, and production for those of ordinary skill having the benefit of this disclosure without undue experimentation.

[0067] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for evaluating the quality of real-time teaching based on brain-computer fusion, characterized in that: The following steps are involved: S1: Based on the international 10-20 lead standard, the consumer-grade EEG acquisition device Emotiv Epoc X dry electrode cap was deployed to collect 14 channels of raw EEG signals from the surface of the brain of the students tested through the dry electrode device, and the data was stored in the form of a two-dimensional array by channel; S2: pre-processing the collected raw EEG signals, removing baseline drift, bandpass filtering and artifacts, and interpolating the detected bad tracks to generate initial processable EEG signals; S3: Extract the differential entropy of the preprocessed EEG signal in frequency bands as the local key signal feature; and stack it into a three-dimensional form as a data set according to the set rules; S4: Input the data set into the pre-built neural network model, adjust the neural network model parameters until the accuracy threshold requirement is met, and propose a teaching quality evaluation model to output the evaluation score results.

2. According to claim 1, a method for evaluating the quality of real-time teaching based on brain-computer fusion is characterized in that: In step S1, the original EEG signals of the students under test are collected based on the international 10-20 lead system standard. The frontal midpoint (Fz), central point (Cz), vertex (Pz), occipital point (Oz), left frontal pole (Fp1), right frontal pole (Fp2), left frontal (F3), right frontal (F4), left central (C3), right central (C4), left temporal middle (T3), right temporal middle (T4), left occipital (O1), and right occipital (O2) are selected as a 14-channel dry electrode device. Two reference sensors are connected to the earlobes in a wet electrode manner to collect the original EEG signals of the students under test. The Bluetooth frequency for collecting the original EEG signals is 2.4 GHz, and the sampling frequency is set to 128 Hz.

3. According to claim 1, a method for evaluating the quality of real-time teaching based on brain-computer fusion is characterized in that: The removal of baseline drift in step S2 is a wavelet transform that decomposes and reconstructs the original EEG signal according to rules to remove redundant signals; the wavelet transform is implemented by the following formula: Where WT(a,τ) represents the signal after wavelet transformation, s(t) is the initial EEG signal, Represents the wavelet basis function; the variables in the wavelet transform include displacement τ and scale a. Scale a determines the analysis accuracy of the wavelet transform in the frequency domain, and the displacement τ determines the position of the wavelet function in the time domain. The wavelet basis function selects Daubechies wavelet to improve the frequency domain localization ability; The filtering described in step S2 uses a zero-phase bandpass filter based on the Butterworth transfer function to filter out the 4 Hz-45 Hz signal. The Butterworth transfer function H(s) is implemented by the following formula: Where ω0 is the center frequency of the filter (set to 25Hz), Q is the filter quality factor, ω s is the angular frequency containing complex variables, K is the actual gain of the filter; the EEG signal of the non-stationary digital signal type is composed of a series of second-order filters; The artifact removal described in step S2 is to decompose the EEG signal into independent components through ICA, identify and remove the artifact components, and finally reconstruct the clean signal and evaluate the removal effect to improve the quality of the EEG signal to obtain an initial processable EEG signal.

4. The method for evaluating the quality of real-time teaching based on brain-computer fusion according to claim 1 is characterized in that: The steps of extracting the differential entropy of the pre-processed EEG signal by frequency bands described in step S3 are as follows: The frequency band feature extraction is achieved through a Butterworth bandpass filter. The EEG signal is decomposed into four frequency bands: theta (4-8Hz), alpha (8-14Hz), beta (14-31Hz) and gamma (31-45Hz). For each channel of each trial, the first 3 seconds are extracted as the baseline signal and the remaining part is extracted as the trial signal. For each frequency band signal (theta, alpha, beta, gamma) after filtering, its differential entropy is calculated. Differential entropy is a measure of signal complexity. It is calculated by calculating the mean square error of the signal and using the following formula: Where e is the natural exponent, σ is the standard deviation of the signal sequence; the calculated two-dimensional data of the frequency-band differential entropy are stacked along the new dimension to form a three-dimensional data set for subsequent classification processing.

5. The method for evaluating the quality of real-time teaching based on brain-computer fusion according to claim 1 is characterized in that: The convolutional neural network model described in step S4 is used to process and classify 4-band EEG signal data, and the model includes in sequence: an input layer for receiving data; a first two-dimensional convolution block, configured as 4 input channels, 24 output channels, a 5*5 convolution kernel and a ReLU activation function; a second two-dimensional convolution block, configured as 24 input channels, 128 output channels, a 2*2 convolution kernel, a ReLU activation function and a 2*2 maximum pooling layer; a first fully connected layer block, with a connection dimension of 128*2*2 input features, an output dimension of 1*96 features and a ReLU activation function; a second fully connected layer block, with a connection dimension of 1*96 input features, an output dimension of 1*2 Features: A Dropout layer with a dropout rate of 0.5 is performed after the second fully connected block to form the forward propagation path of the model to realize the classification processing of the EEG signal data, the loss function in the neural network model is set to cross entropy, the Adam optimization strategy is selected as the optimizer, the learning rate in the optimizer is set to 0.00001, the time series length is set to 8, and the batch_size is set to 32; finally, the three-dimensional test EEG signal data set composed of differential entropy is sent to the training to obtain the neural network model, and the classification test is performed on the Valence and Arousal indicators, and the accuracy threshold is set to 0.8; In order to quantify the teaching quality evaluation results, a teaching quality evaluation method based on Valence and Arousal is proposed. The main formula is as follows: Score=0.4*Valence+0.6*Arousal Among them, the quantization range of Valence is 0-9, and the quantization range of Arousal is 0-9; After reaching the accuracy threshold, the teaching quality evaluation score is given for the EEG signal of the current student under test: the scoring quantitative index range is 0-9, where 0-3.0 is fair, 3.0-6.0 is good, and 6.0-9.0 is excellent.

6. An electronic terminal, characterized in that: including processor and storage medium; The storage medium is used to store instructions; The processor is used to operate according to the instructions to execute the steps of the method according to any one of claims 1 to 5.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images