Brain-computer interface algorithm system based on lower limb motor imagery

Abstract

The invention discloses a brain-computer interface algorithm system based on lower limb motor imagery, relates to the technical field of brain-computer interfaces, and is technically characterized in that a high-precision lower limb motor imagery brain-computer interface system based on self-adaptive time frequency-Riemannian geometry fusion is adopted, and through optimal frequency band selection and covariance manifold classification, a high-precision lower limb motor imagery brain-computer interface algorithm is obtained. Robust decoding of the motion intention of the patient suffering from knee joint pain is achieved, and reliable technical support is provided for clinical rehabilitation.

Description

Technical Field

[0001] The present invention relates to the technical field of brain-computer interface technology, and in particular to a brain-computer interface algorithm system based on lower limb motor imagery. Background Art

[0002] Chronic knee osteoarthritis pain is one of the most common diseases worldwide, significantly impairing patients' quality of life and reducing work productivity. However, its impact on cortical plasticity and motor imagery (MI) remains poorly understood. Although motor imagery-based brain-computer interfaces (MI-BCIs) have made significant progress in upper limb motor function recovery, and MI has been shown to be able to induce neural activity in the motor cortex and achieve high accuracy in electroencephalogram (EEG) signal decoding, its application in lower limb diseases, especially in patients with pain, remains underexplored.

[0003] To this end, the present invention aims to provide a brain-computer interface algorithm system based on lower limb motor imagery to solve the above problems. Summary of the Invention

[0004] The purpose of this invention is to solve the above problems and provide a brain-computer interface algorithm system based on lower limb motor imagery. Through an innovative adaptive time-frequency-Riemannian geometry fusion algorithm, it significantly improves the accuracy of movement intention recognition and provides reliable technical support for clinical rehabilitation.

[0005] In order to achieve the above object, the technical solution of the present invention is as follows:

[0006] The present invention provides a brain-computer interface algorithm system based on lower limb motor imagery, which comprises a data preprocessing module, a time-frequency feature extraction module, a feature dimension reduction module, a geometric filtering and classification module, a decision fusion module and a real-time feedback module;

[0007] The data preprocessing module filters the raw signal collected from the EEG device to remove high-frequency noise and artifacts, such as eye movement artifacts and power supply interference, and outputs the cleaned EEG data signal for further analysis by applying a 0.5 to 80 Hz bandpass filter and a 50 Hz notch filter. The data preprocessing module obtains the clean signal x_f(t) after filtering the raw EEG signal x(t) with a 50 Hz notch filter, and divides x_f(t) into independent trials;

[0008] The time-frequency feature extraction module extracts features from EEG signals using a time-frequency analysis method. Specifically, continuous wavelet transform and Welch method are used to perform frequency domain analysis on EEG signals within different time windows to obtain the energy characteristics of each frequency band (such as α waves and β waves). These time-frequency features can capture the neural activity related to lower limb motor imagery and provide a basis for subsequent feature classification.

[0009] The time-frequency feature extraction module uses the Welch method and continuous wavelet transform to perform time-frequency analysis on the preprocessed EEG signal; the Welch method calculates the power spectral density through the periodogram:

[0010]

[0011] Where w(n) is the Hamming window, L is the window length, and U is the normalization constant; CWT is defined as:

[0012]

[0013] The analysis frequency bands cover 8 intervals: [8,12Hz], [8,20Hz], [8,30Hz], [12,20Hz], [15,20Hz], [15,30Hz], [20,30Hz], and [8,15Hz]. The optimal classification frequency band is determined through adaptive frequency band selection;

[0014] The feature dimensionality reduction module uses a local tangent space arrangement method to perform feature dimensionality reduction on the EEG signal. The high-dimensional features of the EEG signal are compressed to a lower dimension, thereby improving computational efficiency, avoiding overfitting of high-dimensional features, and reducing the risk of overfitting.

[0015] The geometric filtering and classification module uses the minimum distance classification (MDRM) method based on Riemannian geometry to classify the reduced features. By calculating the covariance matrix of different time-frequency windows and using the Riemannian distance to measure the similarity between the signals, the classifier can effectively determine the type of motor imagery (left leg or right leg). The geometric filtering and classification module can robustly classify EEG signals from different individuals, handling differences across individuals and time periods. The Riemannian distance between the two covariance matrices C_i and C_j is:

[0016] δR(Ci,Cj)=|log(Ci -1 / 2 Cj -1 / 2 )| F

[0017] where |·| F is the Frobenius norm, and the Riemann mean C is obtained by minimizing the sum of squared Riemann distances;

[0018] The decision fusion module fuses the classification results of the geometric filtering and classification module with other classification methods (such as time-space linear discriminant analysis (TSLDA)). Decision fusion improves classification accuracy by weighted combination of the outputs of multiple classifiers and can adapt to the EEG patterns of different individuals. The decision fusion module integrates the classification results of time-space linear discriminant analysis and discriminant geodesic filtering (MDRM) and determines the final category through weighted fusion:

[0019]

[0020] Where P_1(c) and P_2(c) are the probability outputs of the two types of devices, and w_1 and w_2 are the corresponding weights;

[0021] The real-time feedback module provides visual or auditory feedback to the user based on the classification results, indicating the execution status of the motor imagery; the user confirms whether the lower limb motor imagery is successful through indicator lights or graphics; the real-time feedback not only helps the user adjust the imagery task, but also promotes the brain's neuroplasticity through the feedback mechanism, which helps to improve the rehabilitation effect.

[0022] The data preprocessing module includes bandpass filtering (0.5-80 Hz) and a bad trial rejection mechanism. When the signal amplitude exceeds ±75 μV in the 8-30 Hz frequency band, it is automatically marked as a bad trial and excluded from the analysis.

[0023] The time-frequency feature extraction module adopts a sliding window strategy to dynamically track the changes in ERD / ERS patterns, with a window length of 1 second and an overlap rate of 50%, to capture the dynamic changes in neural activity during motor imagery.

[0024] When the feature dimension reduction module adopts the local tangent space arrangement algorithm, the first 20 principal components are retained, which can explain more than 90% of the data variance, significantly reducing the computational complexity while ensuring the classification performance.

[0025] When the geometric filtering and classification module adopts the Riemann mean classifier, a specific covariance matrix regularization parameter λ=0.1 is set according to the EEG characteristics of patients with knee pain to improve classification stability.

[0026] In the decision fusion module, the weight w1 of spatiotemporal linear discriminant analysis is set to 0.6, and the weight w2 of discriminant geodesic filtering MDRM is set to 0.4. The weight ratio is determined by cross-validation to obtain the optimal classification performance.

[0027] The system supports cross-session and cross-subject transfer learning. When the pre-trained model is applied to new patients through domain adaptation technology, a classification accuracy of more than 80% can be achieved with only a small amount of calibration data.

[0028] In this scheme, eight sub-bands (α / β and their combinations) are analyzed in parallel by Welch periodogram and continuous wavelet transform (CWT), and the most discriminative joint frequency band [8-30 Hz] is dynamically selected to solve the feature extraction problem caused by the variation of ERD / ERS patterns in pain patients.

[0029] The EEG covariance matrix is mapped to Riemannian space, and the minimum Riemannian distance classifier (MDRM) is used to process non-stationary signals. Compared with the traditional Euclidean space method, the robustness to electrode offset and individual differences is improved by 37.2%.

[0030] The results of spatiotemporal linear discriminant analysis (TSLDA) and Riemannian geometry classification were integrated, and weighted voting (weight ratio 0.6:0.4) was used to balance the complementary information of spatiotemporal features and manifold structure, so that the average accuracy across subjects reached 86.41%.

[0031] A 4-second motor imagery duration and visual-auditory dual-modal prompts are designed for pain patients, combined with a real-time virtual feedback interface, which significantly reduces the invalid trial rate caused by pain distraction (<8%).

[0032] Compared with the existing technology, this solution has the following beneficial effects:

[0033] The present invention validated the EEG dataset by analyzing the proportion of poor trials, event-related desynchronization / synchronization (ERD / ERS) in motor root images, and classification performance; after excluding poor trials, significant contralateral ERD and ipsilateral ERS were observed in the sensorimotor areas (C3 and C4) during the MI task in pain patients, indicating that lower limb MI can significantly enhance the neuroplasticity of the motor areas in pain patients; laying a solid foundation for future clinical applications, especially for individuals with limited lower limb motor function. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 1 is a schematic diagram comparing the classification accuracy of the four models in 30 subjects according to the embodiment of the present invention;

[0035] Figure 2 Schematic diagram of confusion matrix of four classification models in an embodiment of the present invention. DETAILED DESCRIPTION

[0036] In order to enable those skilled in the art to better understand the present invention, the technical solution of the present invention will be further described in detail below in conjunction with the embodiments of the present invention and the accompanying drawings. Obviously, the embodiments described 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 making creative efforts should fall within the scope of protection of the present invention.

[0037] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments of the present invention can be combined with each other. The present invention will be described in detail below with reference to the embodiments.

[0038] Example 1:

[0039] Classification performance verification:

[0040] To rigorously validate a newly collected dataset of lower limb motor imagery in patients with knee pain, this study employed four classification methods—CSP+LDA, FBCSP+SVM, EEGNet, and the newly developed OTFWRGD algorithm—to analyze EEG signal features from different dimensions, reflecting the technological evolution from traditional feature engineering methods to advanced machine learning strategies.

[0041] 1. Traditional spatial filtering method (CSP+LDA)

[0042] Common Spatial Patterns (CSP) is a spatial filtering technique widely used in motor imagery brain-computer interface research. It maximizes the variance difference between two signal types (in this case, left and right lower limb motor imagery) through a linear transformation. Combined with Linear Discriminant Analysis (LDA), CSP+LDA has become a benchmark method for both offline and online brain-computer interface research due to its ease of implementation and strong interpretability. This study used this method to establish a performance baseline, providing a reference for subsequent comparisons of advanced algorithms.

[0043] 2. Extended filter bank method (FBCSP+SVM)

[0044] Filter Bank Common Spatial Patterns (FBCSP) improves upon traditional methods by decomposing EEG signals into multiple frequency bands (α, β, or narrower subbands) and applying CSP to each of these bands. This technique extracts frequency-domain features specific to motor imagery tasks, which are then classified using a support vector machine (SVM). SVMs have proven robust in brain-computer interface applications, and combined with frequency-band-specific spatial filtering, FBCSP+SVM offers unique advantages in assessing the discriminability of cross-band features.

[0045] 3. Deep Learning Architecture (EEGNet)

[0046] EEGNet is a compact deep learning architecture designed specifically for EEG signals. It learns spatiotemporal features directly from raw signals through temporal convolution, spatial convolution, and depthwise separable convolution layers. This study introduced EEGNet to evaluate the ability of end-to-end learning methods to process lower limb motor imagery data from patients with knee osteoarthritis, whose cortical activity patterns may differ from those of healthy individuals. This model, while maintaining a low number of trainable parameters, is adaptable to diverse EEG tasks, making it an ideal tool for validating dataset quality.

[0047] 4. Optimal Time-Frequency Window Riemannian Geometry Classification Algorithm (OTFWRGD)

[0048] The optimal time-frequency window Riemannian geometry distance classification algorithm (OTFWRGD) innovatively integrates time-frequency feature optimization, covariance feature extraction and Riemannian geometry classification methods to improve the classification performance of motor imagery EEG signals through multi-level processing.

[0049] Data preprocessing stage:

[0050] The original EEG signal x(t) is filtered with a 50Hz notch filter to obtain the clean signal x_f(t), which is then divided into independent trials.

[0051] Time-frequency feature extraction:

[0052] The Welch method and continuous wavelet transform (CWT) were used for time-frequency analysis. The Welch method calculates the power spectral density through the periodogram:

[0053]

[0054] Where w(n) is the Hamming window, L is the window length, and U is the normalization constant. The CWT is defined as:

[0055]

[0056] The analysis frequency bands cover 8 intervals including [8,12Hz], [8,20Hz], [8,30Hz], [12,20Hz], [15,20Hz], [15,30Hz], [20,30Hz], and [8,15Hz]. The optimal classification frequency band is determined through adaptive frequency band selection.

[0057] Feature dimensionality reduction:

[0058] The local tangent space alignment (LTSA) algorithm is used to reduce the dimensionality of the covariance matrix, which reduces the computational complexity while preserving the local geometric relationship and avoids overfitting of high-dimensional features.

[0059] Geometric filtering and classification:

[0060] After dimensionality reduction, the features are processed by discriminant geometry filtering and classified using the Riemann geometry method. The Riemann distance between the two covariance matrices C_i and C_j is:

[0061] δR(Ci,Cj)=|log(Ci -1 / 2 Cj -1 / 2 )| F

[0062] where |·| F is the Frobenius norm. The Riemann mean C is obtained by minimizing the sum of squared Riemann distances.

[0063] Decision Fusion:

[0064] The classification results of spatiotemporal linear discriminant analysis (TSLDA) and discriminant geodesic filter MDRM (DGFMDRM) are integrated to determine the final category through weighted fusion:

[0065]

[0066] Among them, P_1(c) and P_2(c) are the probability outputs of the two types of devices, and w_1 and w_2 are the corresponding weights.

[0067] The OTFWRGD algorithm has the following innovative advantages:

[0068] 1. The adaptive optimization mechanism of time-frequency windows improves the ability to capture the neural oscillation characteristics related to motor imagery

[0069] 2. Covariance feature mapping to Riemannian manifold to enhance robustness to artifacts and individual differences

[0070] 3. Amplitude scaling invariance improves model generalization performance

[0071] 4. LTSA dimensionality reduction improves computational efficiency while preserving the intrinsic geometric structure of EEG

[0072] 5. Riemannian distance metric is more adaptable to the characteristics of non-stationary EEG signals than traditional Euclidean method

[0073] 6. Decision fusion strategy improves classification reliability and individual adaptability

[0074] This algorithm provides an innovative solution for motor imagery EEG classification through the organic combination of time-frequency feature optimization selection, covariance feature extraction and Riemannian geometry classification, significantly improving the classification accuracy, computational efficiency and system robustness of brain-computer interface applications.

[0075] Classification performance results:

[0076] To validate the classification performance of the OTFWRGD algorithm, this study conducted a rigorous training and testing process on each subject's EEG data, using a 10-fold cross-validation method to ensure the robustness of the classification results. The following are the detailed steps of this process:

[0077] (1) Dataset division

[0078] First, all EEG data was divided into training and test sets. Each subject's EEG dataset was split in a 7:3 ratio: 70% for training and 30% for testing. This data split ensures fairness between training and testing, while also allowing the test set to fully reflect the model's performance on unseen data.

[0079] (2) Cross-validation

[0080] To further evaluate the robustness of the OTFWRGD algorithm, 10-fold cross-validation was employed. Specifically, the EEG datasets of all subjects were randomly divided into 10 subsets, 9 of which were used for training and the remaining 1 for validation. This process was repeated 10 times, with a different subset selected as the validation set each time. Finally, the average cross-validation accuracy was calculated to avoid evaluation bias caused by accidental data partitioning.

[0081] (3) Model training

[0082] In each fold of cross-validation, the OTFWRGD algorithm is trained using the training set. The training process includes the following steps:

[0083] Data preprocessing: The EEG signals in the training set were subjected to 50 Hz notch filtering and band-pass filtering (0.5 Hz to 80 Hz) to remove artifacts and noise.

[0084] Time-frequency feature extraction: Welch method and continuous wavelet transform (CWT) are used to perform time-frequency analysis on EEG signals and extract time-frequency domain features.

[0085] Feature dimensionality reduction: The local tangent space alignment (LTSA) algorithm is applied to reduce the dimensionality of the time-frequency features to reduce the dimension of the feature space and avoid overfitting.

[0086] Geometric classification: Use Riemannian geometry methods to calculate the Riemannian distance between the covariance matrices after dimensionality reduction to classify the data.

[0087] Decision fusion: Determine the final classification output by weighted fusion of the results of multiple classifiers (such as spatiotemporal linear discriminant analysis TSLDA and discriminant geodesic filtering MDRM).

[0088] (4) Testing and evaluation

[0089] After each training session, the model is evaluated using the test set. During the evaluation process, the system uses the trained model to classify the test set and calculates the following performance metrics:

[0090] Average Precision: measures the proportion of correct classifications performed by the model on the test set.

[0091] Kappa coefficient: Evaluates classification consistency and reliability of the classifier and eliminates the influence of accidental consistency.

[0092] Precision and Sensitivity: Precision reflects the proportion of positive examples correctly predicted by the classifier, and sensitivity reflects the classifier's ability to correctly identify negative examples.

[0093] (5) Classification results

[0094] In this study, each subject's EEG dataset was split into training and test sets in a 7:3 ratio. To further validate model performance, a 10-fold cross-validation method was used to ensure robust estimation of performance metrics across all subjects. Table 2 shows the mean results for the four models (CSP+LDA, FBCSP+SVM, EEGNet, and OTFWRGD) in terms of average accuracy, the Kappa coefficient, a measure of classification consistency, and the precision and sensitivity, robustness indicators of the decoding method. As shown in Table 1, the average accuracy of the four models was 51.43%, 55.71%, 76.21%, and 86.41%, respectively.

[0095] Table 1 Comparison of average classification performance of different models on 30 subjects

[0096]

[0097] Figure 1 The individual classification accuracy of the four models was presented for 30 subjects (a total of 15,000 trials). The results showed that the OTFWRGD method achieved significantly higher accuracy than the other three methods for most subjects. EEGNet performed comparable to or slightly better than traditional methods (CSP+LDA and FBCSP+SVM) in most cases, demonstrating its adaptive feature learning capabilities. However, the accuracy of CSP+LDA and FBCSP+SVM showed greater variability across subjects, suggesting that these methods may be more sensitive to individual differences in knee pain severity or imagery ability.

[0098] To further evaluate the classification performance, Figure 2 The confusion matrices of the four models were presented, and the discrimination of each classifier on the left and right lower limb motor imagery tasks was statistically analyzed. The four classification methods showed differentiated performance characteristics, confirming the applicability of this lower limb motor imagery EEG dataset in the pain patient population. OTFWRGD showed the highest classification accuracy and stability in all subjects, demonstrating its effectiveness in solving the unique challenges of lower limb pain. EEGNet also showed good performance, highlighting the potential of deep learning to automatically extract effective features from clinical EEG data. The results of CSP+LDA and FBCSP+SVM showed that even with traditional machine learning methods, the dataset can still provide reliable motor imagery signals. Overall, these findings confirm the distinguishability of left and right lower limb motor imagery neural activities, providing strong support for the practical value of this dataset in future rehabilitation engineering and brain-computer interface applications.

[0099] Example 2:

[0100] 1. Dataset Overview

[0101] The research team collected a large-scale lower limb motor imagery (MI) electroencephalogram (EEG) dataset, focusing on patients with lower limb knee pain.

[0102] The dataset contains 30 patients, each of whom completed five independent experiments, each of which consisted of 100 left and right leg flexion and extension motor imagery tasks, totaling 15,000 trials.

[0103] The data is organized according to the EEG-BIDS standard, including raw data, preprocessed data, and experimental metadata, and is publicly available on the Figshare platform.

[0104] 2. Experimental Design

[0105] The experiment used a 32-channel semi-dry electrode EEG device (ZhenTec NT1) with a sampling rate of 500 Hz.

[0106] Task design: Each trial consisted of 2 seconds of preparation, 4 seconds of motor imagery, and 4 seconds of rest. Patients were guided to complete motor imagery of their left or right leg through visual cues.

[0107] Data preprocessing included 0.5-80 Hz bandpass filtering, 50 Hz notch filtering, and segmentation and annotation using the EEGLAB toolbox.

[0108] 3. Data Analysis Methods

[0109] ERD / ERS analysis: During motor imagery, significant event-related desynchronization (ERD) and ipsilateral synchronization (ERS) were observed in the contralateral sensorimotor area (C3 / C4 channels), especially in the 8-30 Hz frequency band.

[0110] Classification algorithm:

[0111] Traditional methods: CSP+LDA (51.43% accuracy), FBCSP+SVM (55.71% accuracy).

[0112] Deep learning method: EEGNet (76.21% accuracy).

[0113] New algorithm: OTFWRGD (Optimal Time-Frequency Window in Riemannian Geometric Distance), with an average accuracy of 86.41%, significantly outperforms other methods.

[0114] The OTFWRGD algorithm combines time-frequency feature selection, covariance matrix dimensionality reduction (LTSA) and Riemannian geometry classification (MDRM) to improve performance through decision fusion.

[0115] 4. Key Findings

[0116] Motor imagery in pain patients still induced detectable neural activity, but the magnitude of ERD (-14.6%) was lower than that in healthy controls (-20% to -30%).

[0117] The OTFWRGD algorithm performed well in the classification task, verifying the feasibility of decoding lower limb motor imagery EEG signals.

[0118] The dataset provides an important foundation for the application of BCI in rehabilitation medicine, especially for patients with lower limb movement disorders.

[0119] 5. Limitations and Future Directions

[0120] The sample is limited to patients with KOA and needs to be expanded to other lower limb diseases (such as spinal cord injury) in the future.

[0121] The experimental task design is static, and dynamic scenarios need to be added to improve practicality.

[0122] The real-time performance and computational efficiency of the OTFWRGD algorithm need to be further optimized.

[0123] It is recommended to combine multimodal signals (such as EMG, fNIRS) to improve system robustness.

[0124] 6. Clinical significance

[0125] This confirms that the motor imagery ability of pain patients is preserved, supporting the application of BCI technology in rehabilitation.

[0126] High-precision classification algorithms provide technical support for the development of personalized rehabilitation strategies.

[0127] 7. Abbreviations and terms

[0128] EEG: electroencephalogram; MI: motor imagery; BCI: brain-computer interface; ERD / ERS: event-related desynchronization / synchronization; CSP: common spatial pattern; LDA: linear discriminant analysis; MDRM: minimum distance to the Riemann mean.

[0129] 8. Data Availability

[0130] The dataset and code are publicly available on GigaDB, following the EEG-BIDS standard, to facilitate reuse in subsequent research.

[0131] The above specific embodiments are merely explanations of the present invention and are not limitations of the present invention. After reading this specification, those skilled in the art may make non-creative modifications to the embodiments as needed. However, as long as they are within the scope of the claims of the present invention, they are protected by patent law.

Claims

1. A brain-computer interface algorithm system based on lower limb motor imagery, characterized by: The algorithm system includes a data preprocessing module, a time-frequency feature extraction module, a feature dimension reduction module, a geometric filtering and classification module, a decision fusion module and a real-time feedback module; The data preprocessing module filters the raw signal collected from the EEG device to remove high-frequency noise and artifacts, and outputs the cleaned EEG data signal for further analysis by applying a 0.5 to 80 Hz bandpass filter and a 50 Hz notch filter. The data preprocessing module obtains the clean signal x_f(t) after filtering the raw EEG signal x(t) with a 50 Hz notch filter, and divides x_f(t) into independent trials; The time-frequency feature extraction module performs frequency domain analysis on EEG signals in different time windows using a time-frequency analysis method to obtain energy characteristics of each frequency band; The time-frequency feature extraction module uses the Welch method and continuous wavelet transform to perform time-frequency analysis on the preprocessed EEG signal; the Welch method calculates the power spectral density through the periodogram: Where w(n) is the Hamming window, L is the window length, and U is the normalization constant; CWT is defined as: The analysis frequency bands cover 8 intervals: [8,12Hz], [8,20Hz], [8,30Hz], [12,20Hz], [15,20Hz], [15,30Hz], [20,30Hz], and [8,15Hz]. The optimal classification frequency band is determined through adaptive frequency band selection; The feature dimensionality reduction module uses a local tangent space arrangement method to perform feature dimensionality reduction on the EEG signal, compressing the high-dimensional features of the EEG signal to a lower dimension, thereby improving computational efficiency, avoiding overfitting of high-dimensional features, and reducing the risk of overfitting; The geometric filtering and classification module uses the minimum distance classification method based on Riemannian geometry to classify the features after dimensionality reduction; By calculating the covariance matrices of different time-frequency windows and using the Riemann distance to measure the similarity between signals, the classifier can efficiently determine the type of motor imagery. The geometric filtering and classification module can robustly classify EEG signals from different individuals and handle differences across individuals and time periods. The Riemann distance between the two covariance matrices C_i and C_j is: δR(Ci,Cj)=|log(Ci -1 / 2 CjCi -1 / 2 )| F where |·| F is the Frobenius norm, and the Riemann mean C is obtained by minimizing the sum of squared Riemann distances; The decision fusion module fuses the classification results of the geometric filtering and classification module with the time-space linear discriminant analysis (TSLDA) method. Decision fusion improves classification accuracy by weighted combination of the outputs of multiple classifiers and can adapt to the EEG patterns of different individuals. The decision fusion module integrates the classification results of the time-space linear discriminant analysis and the discriminant geodesic filter (MDRM) and determines the final category through weighted fusion: Where P_1(c) and P_2(c) are the probability outputs of the two types of devices, and w_1 and w_2 are the corresponding weights; The real-time feedback module provides visual or auditory feedback to the user based on the classification results to indicate the execution status of the motor imagery; the user confirms whether the lower limb motor imagery is successful through indicator lights or graphics; the real-time feedback not only helps the user adjust the imagination task.

2. The brain-computer interface algorithm system based on lower limb motor imagery according to claim 1, characterized in that: The data preprocessing module can perform bandpass filtering and a bad trial elimination mechanism. When the amplitude of the signal exceeds ±75μV in the 8-30Hz frequency band, it is automatically marked as a bad trial and excluded from the analysis.

3. The brain-computer interface algorithm system based on lower limb motor imagery according to claim 1, characterized in that: The time-frequency feature extraction module adopts a sliding window strategy to dynamically track the changes in ERD / ERS patterns, with a window length of 1 second and an overlap rate of 50%, to capture the dynamic changes in neural activity during motor imagery.

4. The brain-computer interface algorithm system based on lower limb motor imagery according to claim 1, characterized in that: The feature dimension reduction module adopts the local tangent space arrangement algorithm, retains the first 20 principal components in the local tangent space arrangement algorithm, ensures classification performance, and reduces computational complexity.

5. The brain-computer interface algorithm system based on lower limb motor imagery according to claim 1, characterized in that: When the geometric filtering and classification module adopts the Riemann mean classifier, a specific covariance matrix regularization parameter λ=0.1 is set according to the EEG characteristics of patients with knee pain to improve classification stability.

6. The brain-computer interface algorithm system based on lower limb motor imagery according to claim 1, characterized in that: In the decision fusion module, the weight w1 of spatiotemporal linear discriminant analysis is set to 0.6, and the weight w2 of discriminant geodesic filtering MDRM is set to 0.

4. The weight ratio is determined by cross-validation to obtain the optimal classification performance.

7. The brain-computer interface algorithm system based on lower limb motor imagery according to claim 1, characterized in that: The system can transfer learning across sessions and across subjects.

Images