CKE-based ontology feeling quantification method and CKE-based ontology feeling quantification system

Abstract

The invention discloses a CKE-based proprioceptive quantization method. The method comprises the following steps: synchronously acquiring EEG and ACC signals by using an EEG acquisition system and a micro inertial measurement unit; carrying out purification treatment; adopting continuous wavelet transform (CWT) to extract time-frequency characteristics of EEG and ACC signals, and generating a CKC time-frequency graph; the transfer entropy TE from the acceleration signal to the electroencephalogram signal is calculated, the nonlinear information transfer intensity is described by utilizing probability density distribution, TE values are independently calculated for different frequency components, the frequency domain distribution of TE is generated, and a TE time-frequency graph is formed; generating CKE distribution of a time-frequency domain by using the time-frequency data of the CKC and the TE, finally drawing a CKE time-frequency graph, and intuitively displaying the dynamic characteristics of the ontology sense incoming effect in the time-frequency domain; and outputting a CKE quantification result and an evaluation report. The invention further provides a CKE-based proprioceptive sense quantification system. According to the method, the precision and the real-time performance of the CKE in quantifying the proprioceptive introduction effect are improved.

Description

Technical Field

[0001] This invention belongs to the field of neurorehabilitation engineering and proprioception quantification research, specifically involving a proprioception quantification method and system based on CKE. Background Technology

[0002] Proprioception refers to the human body's ability to perceive the spatial position, direction of movement, and speed of limbs through receptors in muscles, joints, and tendons. It is a crucial foundation for maintaining motor control and postural stability. In recent years, with the development of rehabilitation engineering, neuroscience, and human-computer interaction, how to objectively and quantitatively assess an individual's proprioceptive function has become a research hotspot. Scholars both domestically and internationally have conducted extensive research on proprioceptive measurement methods, encompassing various techniques such as subjective scale assessments, behavioral experiments, and electrophysiological signal analysis.

[0003] With the development of high-precision signal acquisition equipment, proprioceptive measurement methods combining multimodal signals (such as EEG and accelerometer data) have become a trend. This method can reflect the coherence characteristics of the nervous and motor systems, providing the possibility for real-time quantification of proprioception. However, due to the interference of environmental noise and artifacts on EEG and kinematic signals under complex motion conditions, existing feature extraction methods face significant challenges in information directionality analysis and signal synchronization calculation.

[0004] Existing research largely focuses on the coherence analysis of electroencephalogram (EEG) and electromyography (EMG) signals, exploring the interaction between the nervous and motor systems using linear and nonlinear methods. However, this approach still has limitations in capturing the nonlinear and dynamic information flow of sensory input, especially in quantifying the effects of proprioceptive input, where it struggles to fully reflect the complex time-frequency characteristics. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, this invention provides a proprioceptive quantification method and system based on CKE, which can not only reveal the dynamic coupling characteristics of cortical-motor signals, but also provide a new tool for real-time assessment of neuromuscular function and development of personalized rehabilitation plans, further improving the accuracy and real-time performance of CKE in quantifying proprioceptive afferent effects. It has important theoretical value and broad application prospects.

[0006] The technical solution adopted by this invention to solve its technical problem is:

[0007] A proprioceptive quantification method based on CKE includes the following steps:

[0008] Step S1: Simultaneously acquire EEG and ACC signals using the iRecorderW16 EEG acquisition system and the MTw miniature inertial measurement unit (IMU) to ensure high-quality multimodal data. The EEG signal sampling rate is set to 500Hz, and the motion acceleration signal sampling rate is set to 100Hz to ensure the resolution and accuracy of time-domain and frequency-domain analysis.

[0009] Step S2 involves bandpass filtering the EEG signal at 0.5–30 Hz to remove high-frequency noise and baseline drift. Using the EEGLAB toolbox, Independent Component Analysis (ICA) is employed to remove electrooculogram artifacts and other motion interference, and baseline correction is performed to improve the signal-to-noise ratio of the EEG signal. The ACC signal is then low-pass filtered, smoothed, and normalized to retain only the dynamic characteristics within the 0.5–30 Hz range. Simultaneously, a unified timestamp and interpolation method is used to achieve precise synchronization between the EEG and ACC signals, providing consistent input for subsequent joint analysis.

[0010] Step S3: Continuous wavelet transform (CWT) is used to extract the time-frequency features of the EEG and ACC signals. Cortex-motion coherence (CKC) is calculated at different time points within a sliding time window. By adjusting the wavelet scaling parameters and the analysis frequency range, a CKC time-frequency map is generated to characterize the linear coupling features between the two signals.

[0011] Step S4: Calculate the transfer entropy (TE) from the acceleration signal to the EEG signal, and characterize the intensity of nonlinear information transmission using the probability density distribution. Calculate the TE value separately for different frequency components to generate the frequency domain distribution of TE, and extend the TE results to the time dimension of CKC through interpolation and mapping to form a TE time-frequency map;

[0012] Step S5: Using the time-frequency data of CKC and TE, calculate their joint probability distribution p(CKC,TE). Using the kernel density estimation (KDE) method, construct the joint distribution from the values ​​of CKC and TE. Calculate the marginal probability distributions p(CKC) and p(TE) of CKC and TE respectively. Calculate the mutual information value of CKC and TE according to the mutual information formula. Record the mutual information value in each time and frequency window into the CKE matrix to generate the CKE distribution in the time-frequency domain. Finally, draw the CKE time-frequency diagram to intuitively display the dynamic characteristics of proprioceptive input effect in the time-frequency domain.

[0013] Step S6: Output CKE quantification results and evaluation report; including time-frequency visualization unit (generating CKE time-frequency graph) and index analysis unit (statistically analyzing CKE peak value, frequency band distribution and other characteristics under different motion tasks).

[0014] A proprioception quantization system based on CKE includes a modal signal acquisition module, a signal preprocessing module, a CKE quantization analysis module, and a quantization result output module.

[0015] The multimodal signal acquisition module is used to simultaneously acquire EEG signals and ACC signals to obtain multi-source data of cortical activity and proprioceptive input.

[0016] The signal preprocessing module performs filtering, artifact removal, and normalization preprocessing operations on the acquired EEG and ACC signals to improve data quality and analysis accuracy.

[0017] The CKE quantization analysis module is used to calculate the cortical motion entropy CKE based on the preprocessed multimodal signal and extract quantitative indicators that reflect the intensity of proprioceptive input and its cortical response characteristics.

[0018] The quantitative result output module is used to visualize the CKE analysis results or output them as standardized parameters for proprioceptive assessment, rehabilitation feedback, or brain-computer interface system calls.

[0019] The beneficial effects of this invention are mainly reflected in:

[0020] 1. A Cortico-Kinematic Entropy (CKE) index based on CKC and TE was constructed. By introducing the time-frequency domain features of EEG signals and motion signals and combining their coupling information, the proprioceptive afferent effect is quantified, which innovatively solves the limitations of traditional methods.

[0021] 2. By integrating EEG and acceleration signals in a multimodal manner and utilizing Cortico-Kinematic Entropy (CKE) for quantitative analysis of proprioception, we can not only reveal the dynamic coupling characteristics of cortical-motor signals but also provide a new tool for real-time assessment of neuromuscular function and development of personalized rehabilitation programs. Therefore, based on existing research, combining novel signal processing methods and artificial intelligence technologies to further improve the accuracy and real-time performance of CKE in quantifying proprioceptive afferent effects has significant theoretical value and broad application prospects. Attached Figure Description

[0022] Figure 1 This is a flowchart of the process of the method of the present invention.

[0023] Figure 2 This is a time-frequency plot of CKC based on a sliding window.

[0024] Figure 3 The TE time-frequency plot is based on a sliding window, where (a) represents ACC→EEG and (b) represents EEG→ACC.

[0025] Figure 4 This is a time-frequency plot of the CKE results. Detailed Implementation

[0026] The present invention will now be further described with reference to the accompanying drawings.

[0027] Reference Figures 1-4 A proprioceptive quantification method based on CKE includes the following steps:

[0028] Step S1: The iRecorderW16 EEG acquisition system and MTw miniature inertial measurement unit are used to simultaneously acquire EEG signals and ACC signals.

[0029] EEG signal acquisition: The EEG electrodes were placed according to the international standard of 10–20 electrodes, and contact between the EEG electrodes and the scalp was achieved through electrode caps. During the experiment, the subjects performed passive limb movements under proprioceptive stimulation of varying intensities, and EEG and kinematic signals were recorded simultaneously. Leads TP9 and TP10 were connected to the mastoid processes behind the left and right ears, respectively, as reference electrodes. The ground electrode was placed at the center of the top of the head. Motor-related EEG signals were recorded from the C3, C4, and CZ regions selected from the 16-channel EEG setup.

[0030] Kinematic signal acquisition: Kinematic signals are acquired by the MTw miniature inertial measurement unit. The sensor is fixed to the end of the limb and records triaxial acceleration signals, which are time-aligned with the EEG signal system via a wireless synchronization module. An elastic strap is used to fix the sensor in place to minimize signal artifacts caused by relative slippage during measurement.

[0031] Step S2 involves bandpass filtering the EEG signal in the range of 0.5–30 Hz, using the EEGLAB toolbox, and removing electrooculogram artifacts and other motion interferences through independent component analysis (ICA), followed by baseline correction. The ACC signal is then subjected to low-pass filtering, smoothing, and norm normalization, retaining only the dynamic characteristics within the 0.5–30 Hz range. Simultaneously, a unified timestamp and interpolation method is used to achieve precise synchronization between the EEG and ACC signals, providing consistent input for subsequent joint analysis.

[0032] The acquired acceleration signals were processed using MATLAB software through low-pass filtering, smoothing, and vector norm normalization to obtain a uniform representation of motion amplitude. The EEGLAB toolbox was then used to remove baseline drift, eye movement, high-frequency noise, and power line interference from the EEG signals.

[0033] Step S3: Continuous Wavelet Transform (CWT) is used to perform time-frequency decomposition on the EEG and ACC signals to obtain the energy distribution of the two time series in the time-frequency domain. Taking the preprocessed measured data from Step S2 to construct two time series of EEG signal x(t) and acceleration signal y(t) as an example, Morlet wavelet function is used for multi-scale analysis to construct the wavelet basis function, as shown in the following formula:

[0034]

[0035] Where ψ(t) is the Morlet mother wavelet, a is the scaling factor, and b is the time translation parameter.

[0036] Within each time window (length L, step size S), the CWT coefficients are locally smoothed to estimate the cross-spectrum and auto-spectrum:

[0037]

[0038]

[0039] Where W represents the sample set within the sliding time window. This leads to the time-frequency cortical-motion coherence (CKC):

[0040]

[0041] By adjusting wavelet parameters (such as scale density and mother wavelet center frequency) and frequency range (such as 2–45Hz), the CKC time-frequency plot can be drawn.

[0042] Step S4: To obtain TE in the time-frequency domain, the signal is first decomposed into frequency bands based on CWT, and B is processed for each frequency band. k Calculate the TE separately. Taking y→x as an example, let the embedding dimensions be k and l, and the delay be τ. Then, the TE over the time window W is defined as:

[0043]

[0044] in

[0045]

[0046] Local window W at each (t,f) t,f Internal Computation TE y→x (t,f) directly yields the TE time-frequency diagram.

[0047] Step S5: Using the time-frequency data of CKC and TE, calculate their joint probability distribution p(CKC,TE). Using the kernel density estimation (KDE) method, construct the joint distribution from the values ​​of CKC and TE. Calculate the marginal probability distributions p(CKC) and p(TE) of CKC and TE respectively. Calculate the mutual information value of CKC and TE according to the mutual information formula. Record the mutual information value in each time and frequency window into the CKE matrix to generate the CKE distribution in the time-frequency domain. Finally, draw the CKE time-frequency diagram to intuitively display the dynamic characteristics of proprioceptive input effect in the time-frequency domain.

[0048] After completing the CKC time-frequency plot calculation in step S4 and the TE time-frequency plot calculation in step S3, two-dimensional kernel density estimation is required to estimate the joint probability distribution p(C,E) and the marginal probability distributions p(C) and p(E) at each time-frequency window (t,f):

[0049]

[0050] Among them (c i ,e i ) represents a sample pair, K(·) represents a two-dimensional Gaussian kernel, and H represents the bandwidth matrix.

[0051] Using the time-frequency information of CKC and TE, a cortical motion entropy (CKE) index is constructed and the quantification result is output. Cortical motion entropy (CKE) is defined as the mutual information between CKC and TE.

[0052]

[0053] Store CKE(t,f) at each time point and frequency point in a matrix CKE∈R. T×F In this process, a time-frequency domain CKE distribution is formed, which intuitively demonstrates the dynamic characteristics of the proprioceptive afferent effect.

[0054] Step S6: Output CKE quantization results and evaluation report; including generating CKE time-frequency plots and statistically analyzing CKE peak values ​​and frequency band distribution characteristics under different motion tasks.

[0055] It includes a time-frequency visualization unit (generating CKE time-frequency graphs) and an index analysis unit (statistically analyzing CKE peak values, frequency band distributions, and other characteristics under different motion tasks).

[0056] Based on the above indicators, and combining simulation and experimental data, a closed-loop control algorithm with CKE as its core is designed (e.g., Figure 1 As shown, the rhythm and intensity of rehabilitation training movements are adjusted based on the real-time calculated CKE value. A deep learning model is used to analyze the dynamic changes of CKE, predict the patient's motor state during the motor task, and provide personalized training suggestions. The CKE calculation efficiency is optimized and embedded into the brain-computer interface to achieve real-time closed-loop control.

[0057] A proprioceptive quantization system based on CKE includes a multimodal signal acquisition module, a signal preprocessing module, a CKE quantization analysis module, and a quantization result output module.

[0058] The multimodal signal acquisition module is used to simultaneously acquire electroencephalogram (EEG) signals and accelerometer (ACC) signals to obtain multi-source data of cortical activity and proprioceptive input.

[0059] The signal preprocessing module performs preprocessing operations such as filtering, artifact removal, and normalization on the acquired EEG and ACC signals to improve data quality and analysis accuracy.

[0060] The CKE quantization analysis module is used to calculate the cortical motion entropy (CKE) based on the preprocessed multimodal signals and extract quantitative indicators that reflect the intensity of proprioceptive input and its cortical response characteristics.

[0061] The quantitative result output module is used to visualize the CKE analysis results or output them as standardized parameters for proprioceptive assessment, rehabilitation feedback, or brain-computer interface system calls.

[0062] Experimental verification and effect evaluation

[0063] 1) Healthy human experiment: EEG and ACC signals were collected using a self-developed multi-mode proprioception assessment device to verify the performance of the signal acquisition equipment and the accuracy of the CKE calculation method.

[0064] 2) Patient experiment: Design actual rehabilitation tasks for stroke patients, collaborate with rehabilitation physicians to determine movement requirements and experimental paradigms, and observe the changes in CKE indicators under different task conditions.

[0065] Based on the dynamic trend of CKE, its effectiveness as a quantitative indicator of proprioceptive input effect is evaluated. Rehabilitation effects are assessed using patient motor function scores (such as the Fugl-Meyer score). CKE values ​​under different rehabilitation task conditions are compared to verify its sensitivity to proprioceptive input. The effects of closed-loop control and traditional rehabilitation training are compared to verify the system's role in improving the patient's rehabilitation process.

[0066] The effectiveness of proprioception depends on the synchronicity of cortical-motor signals (characterized by CKC) and the directionality and intensity of information transmission (characterized by TE). CKE characterizes the interdependence between CKC and TE by calculating their mutual information, thus providing a more comprehensive reflection of the dynamic characteristics of proprioception.

[0067] Experimental Data: EEG and acceleration data under different tasks have been collected. The reliability of the data was ensured through multiple experiments, and preliminary statistical analysis was conducted, yielding meaningful results. The results are as follows: Figures 2-4 As shown.

[0068] The embodiments described in this specification are merely examples of implementations of the inventive concept and are for illustrative purposes only. The scope of protection of this invention should not be considered limited to the specific forms described in these embodiments; rather, it extends to equivalent technical means conceived by those skilled in the art based on the inventive concept.

Claims

1. A proprioceptive quantification method based on CKE, characterized in that, The method includes the following steps: Step S1: Simultaneously acquire EEG and ACC signals using an EEG acquisition system and a miniature inertial measurement unit; Step S2 involves bandpass filtering the EEG signal in the range of 0.5–30 Hz, using the EEGLAB toolbox, and removing electrooculogram artifacts and other motion interferences through independent component analysis (ICA), followed by baseline correction. The ACC signal is then subjected to low-pass filtering, smoothing, and norm normalization, retaining only the dynamic characteristics within the 0.5–30 Hz range. Simultaneously, a unified timestamp and interpolation method is used to achieve precise synchronization between the EEG and ACC signals, providing consistent input for subsequent joint analysis. Step S3: Continuous wavelet transform (CWT) is used to extract the time-frequency features of EEG and ACC signals. Cortical-motion coherence (CKC) at different time points is calculated within a sliding time window. By adjusting the wavelet scaling parameters and the analysis frequency range, a CKC time-frequency diagram is generated to characterize the linear coupling features between the two. Step S4: Calculate the transfer entropy TE from the acceleration signal to the EEG signal, use the probability density distribution to characterize the nonlinear information transmission intensity, calculate the TE value separately for different frequency components, generate the frequency domain distribution of TE, and extend the TE result to the time dimension of CKC through interpolation and mapping to form the TE time-frequency map. Step S5: Using the time-frequency data of CKC and TE, calculate their joint probability distribution p(CKC,TE). Using the kernel density estimation (KDE) method, construct the joint distribution from the values ​​of CKC and TE. Calculate the marginal probability distributions p(CKC) and p(TE) of CKC and TE respectively. Calculate the mutual information value of CKC and TE according to the mutual information formula. Record the mutual information value in each time and frequency window into the CKE matrix to generate the CKE distribution in the time-frequency domain. Finally, draw the CKE time-frequency diagram to intuitively display the dynamic characteristics of proprioceptive input effect in the time-frequency domain. Step S6: Output CKE quantization results and evaluation report; including generating CKE time-frequency graphs and statistically analyzing CKE peak values ​​and frequency band distribution characteristics under different motion tasks.

2. The proprioceptive quantification method based on CKE as described in claim 1, characterized in that, In step S1, the iRecorderW16 EEG acquisition system is used to acquire EEG signals. The EEG electrodes are arranged in accordance with the international standard of 10–20 electrode placement. The EEG electrodes are made in contact with the scalp through electrode caps. Passive limb movements are performed under proprioceptive stimulation of different intensities, and EEG signals and kinematic signals are recorded synchronously. Leads TP9 and TP10 are connected to the mastoid processes behind the left and right ears of the subject as reference electrodes, respectively. The ground electrode is placed in the center of the top of the head. The C3, C4 and CZ areas are selected from the 16-lead EEG channels to record the corresponding movement-related EEG signals.

3. The proprioceptive quantification method based on CKE as described in claim 1, characterized in that, In step S1, an MTw miniature inertial measurement unit is used to collect acceleration signals. The sensor is fixed to the end of the limb to record triaxial acceleration signals and is time-aligned with the EEG signal system through a wireless synchronization module. The sensor is fixed with an elastic strap to ensure that signal artifacts caused by relative slippage are reduced during the measurement process.

4. A proprioceptive quantification method based on CKE as described in any one of claims 1 to 3, characterized in that, In step S3, continuous wavelet transform (CWT) is used to perform time-frequency decomposition on the EEG and ACC signals to obtain the energy distribution of the two time series in the time-frequency domain. The EEG signal x(t) and acceleration signal y(t) are constructed using the preprocessed measured data from step S2. Multi-scale analysis is then performed using the Morlet wavelet function to construct the wavelet basis function, as shown in the following formula: Where ψ(t) is the Morlet mother wavelet, a is the scale factor, and b is the time shift parameter; Within each time window, the length L of the time window and the step size S of the time window are used to locally smooth the CWT coefficients and estimate the cross-spectrum and auto-spectrum: Where W represents the set of samples within the sliding time window, thus obtaining the time-frequency cortical-motion coherence CKC: By adjusting the wavelet parameters and frequency range, the CKC time-frequency diagram is plotted.

5. A proprioceptive quantification method based on CKE as described in any one of claims 1 to 3, characterized in that, In step S4, to obtain TE in the time-frequency domain, the signal is first decomposed into frequency bands based on CWT, and B is processed for each frequency band. k Calculate TE separately. Taking y→x as an example, let the embedding dimensions be k and l, and the delay be τ. Then, TE is defined as follows over the time window W: in Local window W at each (t,f) t,f Internal Computation TE y→x (t,f) directly yields the TE time-frequency diagram.

6. The proprioceptive quantification method based on CKE as described in claim 4, characterized in that, In step S5, two-dimensional kernel density estimation is used to estimate the joint probability distribution p(C,E) and the marginal probability distributions p(C) and p(E) at each time-frequency window (t,f): Among them (c i ,e i ) represents a sample pair, K(·) represents a two-dimensional Gaussian kernel, and H represents the bandwidth matrix; Using the time-frequency information of CKC and TE, a cortical motion entropy CKE index is constructed and the quantification result is output. The cortical motion entropy CKE is defined as the mutual information between CKC and TE. Store CKE(t,f) at each time point and frequency point in a matrix CKE∈R. T×F In this process, a time-frequency domain CKE distribution is formed, which intuitively demonstrates the dynamic characteristics of the proprioceptive afferent effect.

7. A system implemented using the CKE-based proprioception quantization method as described in claim 1, characterized in that, The system includes a modal signal acquisition module, a signal preprocessing module, a CKE quantization analysis module, and a quantization result output module; The multimodal signal acquisition module is used to simultaneously acquire EEG signals and ACC signals to obtain multi-source data of cortical activity and proprioceptive input. The signal preprocessing module performs filtering, artifact removal, and normalization preprocessing operations on the acquired EEG and ACC signals to improve data quality and analysis accuracy. The CKE quantization analysis module is used to calculate the cortical motion entropy CKE based on the preprocessed multimodal signal and extract quantitative indicators that reflect the intensity of proprioceptive input and its cortical response characteristics. The quantitative result output module is used to visualize the CKE analysis results or output them as standardized parameters for proprioceptive assessment, rehabilitation feedback, or brain-computer interface system calls.

Images