A method for discriminating different motor body ownership awareness based on electroencephalogram signals

Abstract

The present invention relates to a method for discriminating different motor body ownership awareness based on electroencephalogram (EEG) signals, which collects the EEG signals of the subject in a static state and during actual leg-lifting movements; collects the EEG signals of the subject under illusory conditions and observation conditions; filters the EEG signals into seven frequency bands (δ: 0.5 - 3.5 Hz, θ: 4 - 7.5 Hz, α: 8 - 11 Hz, β1: 12 - 15.5 Hz, β2: 16 - 19.5 Hz, γ1: 20 - 33.5 Hz, γ2: 34 - 98.5 Hz), calculates the power spectral density (PSD) of the EEG signals in different frequency bands when the subject is in a static state, under observation conditions, under illusory conditions and during actual movements; calculates the partial directed coherence values of the EEG signals when the subject is in a static state, under observation conditions, under illusory conditions and during actual movements, and obtains a PDC connection matrix; fuses the features of the PDC and the PSD to obtain a new index PPDC, which is used to evaluate whether the subject generates body ownership awareness during visual motor illusion; and verifies the feasibility of the PPDC.

Description

Technical Field

[0001] The present invention relates to a method for discriminating different motor body ownership awareness based on electroencephalogram signals. The method for discriminating different motor body ownership awareness based on electroencephalogram signals is a quantization method for inducing body ownership awareness of lower limb visual motor illusion based on electroencephalogram signals, and belongs to the technical field of bioelectrical signal processing. Background Art

[0002] The feeling that a participant has a series of movements without actually acting by himself is called kinesthetic illusion. Kinesthetic illusion is caused by various sensory inputs and can be induced by stimulating the sensory system, such as isometric muscle vibration, skin stretching touch, and vision. Kaneko et al. reported that when a virtual body part is set above an actual body part, visual stimulation using repeated self-motion images causes kinesthetic illusion; this phenomenon is called visual motor illusion (VMI). Visual motor illusion is a subjective experience in which the body does not perform actual movements, but the brain experiences an illusion of body movement, which is caused by watching the participant's own movement video. Visual stimulation produces an illusion of limb movement.

[0003] Studies have shown that the induction of visual motor illusion changes the functional connectivity of the brain's resting state. The activity of the brain is also related to visual motor illusion. Ambron et al. found that magnifying the view of the hand increases the excitability of the primary motor area (M1). That is, kinesthetic illusion increases cortical nerve excitability, especially in the illusion state, and for the muscle groups corresponding to the direction of the movement illusion, regardless of the induction method (visual or tendon vibration), the kinesthetic illusion that induces the feeling of limb movement increases the excitability of cortical nerve cells. Past studies have supported that this change is caused by illusion (not just sensory input). During wrist dorsiflexion and palmar flexion VMI, functional magnetic resonance imaging (FMRI) measured the activities in areas such as the premotor cortex and the inferior parietal lobule. The influence of illusory movement on the excitability of cortical nerve cells is not limited to the upper limb, but also extends to the lower limb. During VIM of ankle dorsiflexion or plantar flexion, even the occurrence of movement illusion in the lower limb was observed, accompanied by an increase in cortical motor excitability, which is similar to that observed in the upper limb.

[0004] As a special visual induced illusion, mirror illusion has been widely studied in recent years and is considered as a tool to promote the motor rehabilitation of hemiplegic patients. By watching the movement of the healthy hand in the mirror, they can feel the movement of their paralyzed or even non-existent hand. Due to upper limb VMI, especially elbow flexion and extension in stroke hemiplegic patients, the range of joint movement and muscle activity has been improved. The study on stroke hemiplegic patients with ankle dorsiflexion dysfunction confirmed that VMI improved the automatic range of ankle dorsiflexion and increased the maximum walking speed. These results all confirmed that the kinesthetic illusion generated by watching the image of self-movement in the video can improve motor function.

[0005] Optokinetic illusion can be regarded as a system that uses its own visual feedback to enhance reality, which can induce a sense of body ownership and activation of the brain regions related to movement. By watching the image of one's own movement, optokinetic illusion can be induced, and at the same time, a sense of body ownership is evoked.

[0006] Currently, a non-invasive scalp electroencephalogram (EEG), motor output type brain-machine interface (BMI) has been applied in the rehabilitation process of stroke hemiplegic patients. Shindo et al. studied the effect of BMI training on the motor function of stroke and moderate to severe hemiplegic patients. During BMI training, patients imagined the extension of their fingers, and during this process, the electroencephalogram recorded the activity of the sensorimotor cortex. This finding indicated that BMI training was effective for the recovery of motor function in stroke hemiplegic patients. Motor imagination is difficult to perform in some patients with severe stroke symptoms, hemiplegic patients, and some elderly people with limited mobility. Some activities generated by VMI in the brain network are similar to those generated during motor execution, thus passively triggering motor imagination. Kaneko et al. showed that VMI can make BMI training feasible. Summary of the Invention

[0007] To overcome the deficiencies of existing research, the present invention provides a method for discriminating the degree of body ownership based on electroencephalogram (EEG) signals. EEG signals of subjects at rest and during actual leg-lifting movements were collected; EEG signals of subjects under illusion conditions and observation conditions were collected; the EEG signals were filtered into seven frequency bands (δ: 0.5 - 3.5 Hz, θ: 4 - 7.5 Hz, α: 8 - 11 Hz, β1: 12 - 15.5 Hz, β2: 16 - 19.5 Hz, γ1: 20 - 33.5 Hz, γ2: 34 - 98.5 Hz), and the power spectral density (PSD) of the EEG signals in different frequency bands was calculated for the subjects at rest, under observation conditions, under illusion conditions, and during actual movements; the partial directed coherence values of the EEG signals of the subjects at rest, under observation conditions, under illusion conditions, and during actual movements were calculated to obtain a PDC connection matrix; the PDC and PSD were subjected to feature fusion to obtain a new index PPDC, which was used to evaluate whether the subjects had a sense of body ownership during visual motion illusion; the feasibility of the PPDC was verified.

[0008] The specific steps of a method for discriminating the sense of body ownership of different movements based on EEG signals are as follows:

[0009] Step 1: Before the experiment, collect the EEG signals of the subjects at rest and during actual leg-lifting movements.

[0010] Step 2: Collect the EEG signals of the subjects under observation conditions and illusion conditions.

[0011] Step 2.1: Collect the EEG signals of the subjects under illusion conditions.

[0012] Step 2.2: Collect the EEG signals of the subjects under observation conditions.

[0013] Step 3: Preprocess the EEG signals obtained in Step 1 and Step 2.

[0014] Step 4: Calculate the power spectral density (PSD) of the preprocessed EEG signals in Step 3.

[0015] Step 5: Calculate the partial directed coherence values of the preprocessed EEG signals in Step 3 to obtain a PDC connection matrix.

[0016] Step 6: We perform feature fusion on the PSD obtained in Step 4 and the PDC obtained in Step 5 to obtain a new index PPDC to estimate whether the subjects have a sense of body ownership during visual motion illusion.

[0017] Step 6.1: We average the power spectral density values in each frequency range and represent the power spectral density of each electrode in the form of a matrix:

[0018]

[0019] Among them, P i Represents the average value of the power spectral density corresponding to the i-th electrode.

[0020] Step 6.2: Model the EEG signal using a multivariate autoregressive model. Use the AIC criterion to determine the order of the model. The directed information flow from channel j to channel i at frequency f, i.e., the PDC value, can be solved using the following formula:

[0021]

[0022] Then the representation of the PDC matrix is:

[0023]

[0024] Among them, pdc ij Refers to the PDC value between the i-th electrode and the j-th electrode. The matrix is a 39×39 two-dimensional square matrix.

[0025] Step 6.3: We multiply the matrix P obtained from PSD with the PDC connection matrix to obtain the matrix PPDC:

[0026]

[0027] Step 7: We input the PPDC matrix obtained in step 6 into the convolutional neural network to identify different motion conditions and verify the feasibility of PPDC.

[0028] Step 1: The first set of experiments was to measure the classification accuracy under seven different experimental conditions (static state, observation condition with a leg-lifting angle of 30 degrees, observation condition with a leg-lifting angle of 60 degrees, illusion condition with a leg-lifting angle of 30 degrees, illusion condition with a leg-lifting angle of 60 degrees, and control experiment with leg-lifting angles of 30 degrees and 60 degrees).

[0029] Step 2: The second set of experiments was to classify the accuracy of the observation condition and the illusion condition at the same leg-lifting angle.

[0030] Furthermore, in step 1, a 64-channel wireless EEG system (NeuSen.W64, Neuracle, China) was used to collect EEG data at a sampling frequency of 1000 Hz. Based on the international 10-20 standard, 39 of the 64 EEG channels were selected for measurement. Before data collection, the impedance was maintained below 5 kΩ by injecting conductive gel. EEG data were collected from each subject both at rest and during the actual kicking exercise.

[0031] Further, the illusion conditions in step two include videos of the subject performing sitting front kicks with the left and right legs at leg-lifting angles of 30 degrees and 60 degrees respectively. To give the subject a more realistic illusion, we recorded videos of the subject performing sitting front kicks with the left and right legs at leg-lifting angles of 30 degrees and 60 degrees respectively one week before the experiment. We imported the recorded leg movement videos into a VR (virtual reality headset), adjusted their positions to create the illusion that the subject's own calves were moving. The observation condition used the same movement videos as the illusion condition. The video was placed directly in front of the subject to observe both the video and their own legs simultaneously.

[0032] Further, in step three, the EEGLAB toolbox in MATLAB was used to preprocess the collected EEG signals. The specific processing steps are as follows:

[0033] Step 3.1: Locate the channels according to the international 10-20 standard.

[0034] Step 3.2: Perform channel selection. Thirty-nine channels were selected for measurement from 64 electroencephalogram channels. Since the outermost channels are greatly affected by eye movements and sounds, they were excluded.

[0035] Step 3.3: Reduce the sampling rate. The sampling rate of the EEG signal was reduced to 250 Hz.

[0036] Step 3.4: Denoise. Cleanline (an EEGLAB plugin) was used to remove sinusoidal artifacts in the scalp channels that were not effectively removed by notch filtering, and other artifacts were removed by independent component analysis (ICA).

[0037] Step 3.5: Filter. The denoised EEG signal was band-pass filtered using the FIR digital filter in EEGLAB and divided into seven frequency bands (delta: 0.5 - 3.5 Hz, theta: 4 - 7.5 Hz, aleph: 8 - 11.5 Hz, beta1: 12 - 15.5 Hz, beta2: 16 - 19.5 Hz, gram1: 20 - 33.5 Hz, gram2: 34 - 98.5 Hz).

[0038] Step 3.6: Re-reference the signal.

[0039] Further, in step four, the power spectral densities of the subject's left and right legs were calculated for seven different movement conditions, namely the static state, the observation condition with a kicking angle of 30 degrees, the observation condition with a kicking angle of 60 degrees, the illusion condition with a kicking angle of 30 degrees, the illusion condition with a kicking angle of 60 degrees, the actual kicking movement with a kicking angle of 30 degrees, and the actual kicking movement with a kicking angle of 60 degrees, at different frequency bands.

[0040] Furthermore, Step 5 is based on Step 4, showing that there are significant differences between different movement conditions in the gram2 frequency band. The PDC connection matrices of the subject's left and right legs in seven different movement conditions in the gram2 frequency band were calculated. To better understand the PDC connection matrix, we calculated the clustering coefficient and global efficiency.

[0041] Furthermore, Step 6 mainly fuses the P matrix and PDC matrix obtained in Steps 4 and 5 to obtain the PPDC to measure the changes in the cerebral cortex of the subject during the static state, observation condition, illusion condition, and actual kicking movement. Whether the subject has a movement intention is estimated according to the value of the PPDC.

[0042] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0043] In previous work, whether the sense of body ownership is generated was only judged subjectively by the subject, and there was no objective experimental study to prove it. Previous studies have shown that the parietal cortex, premotor cortex, intraparietal sulcus, and insula of the brain contribute to the generation of the sense of body ownership. It has been proposed that the brain has an internal prediction model, based on which an efferent copy of a movement command is created and compared with the actual sensory input generated by the movement. If the efferent copy matches the incoming input (proprioceptive input from the moving hand, visual input of the moving object), then the movement is considered to be caused by oneself and the sense of body ownership is generated. We can quantify whether the sense of body ownership is generated by the magnitude of the PPDC value in different regions of the brain. The PPDC can not only be used to display the biomarker of the cerebral cortex change caused by VMI, but also be used as a quantification tool for whether the sense of body ownership is generated. This result is of great significance for the rehabilitation treatment of stroke patients in the future. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0045] Figure 1 It is the experimental flowchart of the method for discriminating different movement body ownership awareness based on EEG signals of the present invention;

[0046] Figure 2 It is the experimental paradigm and the order of each experiment of the present invention;

[0047] Figure 3 It is the power spectral density (PSD) of seven frequency bands of the EEG signals of the subject shown in the present invention under seven different movement conditions;

[0048] Figure 4 The PDC connection matrix obtained from the EEG signals of the subjects shown in the present invention under seven different movement conditions in the gram2 frequency band;

[0049] Figure 5 The analysis index of the PDC connection matrix shown in the present invention;

[0050] Figure 6 The PPDC matrix under different movement conditions of the present invention. Detailed implementation manners

[0051] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0052] The specific steps of a method for discriminating different movement body ownership awareness based on EEG signals are as follows:

[0053] Step 1: Collect the EEG signals of the subjects in the static state and during actual leg-lifting movements before the experiment starts.

[0054] Step 2: Collect the EEG signals of the subjects when they are in the observation condition and the illusion condition.

[0055] Step 2.1: Collect the EEG signals of the subjects when they are in the illusion condition.

[0056] Step 2.2: Collect the EEG signals of the subjects when they are in the observation condition.

[0057] Step 3: Preprocess the EEG signals obtained in Step 1 and Step 2.

[0058] Step 4: Calculate the power spectral density (PSD) of the EEG signals preprocessed in Step 3.

[0059] Step 5: Calculate the partial directed coherence values of the EEG signals preprocessed in Step 3 to obtain the PDC connection matrix.

[0060] Step 6: We perform feature fusion on the PSD obtained in Step 4 and the PDC obtained in Step 5 to obtain a new index PPDC to estimate whether the subjects generate body ownership awareness during visual motion illusion.

[0061] Step 6.1: We average the power spectral density values in each frequency range and represent the power spectral density of each electrode in the form of a matrix:

[0062]

[0063] Among them, P i represents the average value of the power spectral density corresponding to the i-th electrode.

[0064] Step 6.2: Model the EEG signals using a multivariate autoregressive model, and determine the order of the model using the AIC criterion. Then, the directed information flow from channel j to channel i at frequency f, i.e., the PDC value, can be solved using the following formula:

[0065]

[0066] The representation form of the PDC matrix is:

[0067]

[0068] Among them, pdc ij refers to the PDC value between the i-th electrode and the j-th electrode, and the matrix is a 39×39 two-dimensional square matrix.

[0069] Step 6.3: We multiply the matrix P obtained from PSD and the PDC connection matrix to obtain the matrix PPDC:

[0070]

[0071] Step Seven: We input the PPDC matrix obtained in Step Six into a convolutional neural network for the recognition of different movement conditions to verify the feasibility of PPDC.

[0072] Step1: The first set of experiments is on the classification accuracy of seven different experimental conditions (rest state, observation condition with a 30-degree leg lift, observation condition with a 60-degree leg lift, illusion condition with a 30-degree leg lift, illusion condition with a 60-degree leg lift, control experiments with 30-degree and 60-degree leg lifts);

[0073] Step2: The second set of experiments is on the classification accuracy of the observation condition and the illusion condition at the same leg lift angle.

[0074] Furthermore, in Step One, EEG data is collected using a 64-channel wireless EEG system (NeuSen.W64, Neuracle, China) with a sampling frequency of 1000 HZ. Referring to the international 10-20 standard, 39 channels are selected from 64 EEG channels for measurement. Before data collection, the impedance is maintained below 5 KΩ by injecting conductive gel. EEG data of each subject in the rest state and actual kicking movements is collected.

[0075] Further, the illusion conditions in step two include videos of the subject performing sitting front kicks with the left and right legs at leg-lifting angles of 30 degrees and 60 degrees respectively. To give the subject a more realistic illusion, we recorded videos of the subject performing sitting front kicks with the left and right legs at leg-lifting angles of 30 degrees and 60 degrees respectively one week before the experiment. We imported the recorded leg movement videos into a VR (virtual reality headset), adjusted their positions to create the illusion that the subject's own calves were moving. The observation condition used the same movement videos as the illusion condition. The video was placed directly in front of the subject to observe both the video and their own legs simultaneously.

[0076] Further, in step three, the EEGLAB toolbox in MATLAB was used to preprocess the collected EEG signals. The specific processing steps are as follows:

[0077] Step 3.1: Locate the channels according to the international 10-20 standard.

[0078] Step 3.2: Perform channel selection. Thirty-nine channels were selected for measurement from 64 EEG channels. Since the outermost channels are greatly affected by eye movements and sounds, they were excluded.

[0079] Step 3.3: Reduce the sampling rate. The sampling rate of the EEG signal was reduced to 250 Hz.

[0080] Step 3.4: Denoise. Cleanline (an EEGLAB plugin) was used to remove sinusoidal artifacts in the scalp channels that were not effectively removed by notch filtering, and other artifacts were removed by independent component analysis (ICA).

[0081] Step 3.5: Filter. The denoised EEG signal was band-pass filtered using the FIR digital filter in EEGLAB and divided into seven frequency bands (delta: 0.5 - 3.5 Hz, theta: 4 - 7.5 Hz, aleph: 8 - 11.5 Hz, beta1: 12 - 15.5 Hz, beta2: 16 - 19.5 Hz, gram1: 20 - 33.5 Hz, gram2: 34 - 98.5 Hz).

[0082] Step 3.6: Re-reference the signal.

[0083] Further, in step four, the power spectral densities of the subject's left and right legs were calculated for seven different movement conditions at different frequency bands, namely the static state, the observation condition with a kicking angle of 30 degrees, the observation condition with a kicking angle of 60 degrees, the illusion condition with a kicking angle of 30 degrees, the illusion condition with a kicking angle of 60 degrees, the actual kicking movement with a kicking angle of 30 degrees, and the actual kicking movement with a kicking angle of 60 degrees.

[0084] Further, Step 5 is based on Step 4, and there are significant differences between different motion conditions in the gram2 frequency band. Calculate the PDC connection matrix of the left and right legs of the subject in seven different motion conditions in the gram2 frequency band. To better understand the PDC connection matrix, we calculated the clustering coefficient and global efficiency.

[0085] Further, Step 6 mainly fuses the P matrix and PDC matrix obtained in Steps 4 and 5 to obtain the PPDC to measure the changes in the cerebral cortex of the subject during the static state, observation condition, illusion condition, and actual kicking movement. Estimate whether the subject has a movement intention based on the value of the PPDC.

[0086] Embodiment

[0087] Step 1: Obtain the human electroencephalogram (EEG) signal sample data. First, collect the EEG signals of the human body in the static state, illusion condition, observation condition, and actual kicking movement through an EEG signal collector; the specific operation process is as follows:

[0088] Twelve healthy subjects (8 males and 4 females; aged 20 - 33 years; height 165 - 178 cm, weight 52 - 77 kg) were selected for the experiment. The experiment was carried out in an electrically shielded room. The experiment mainly consisted of two parts: the illusion condition and the observation condition. During the whole experiment process, the EEG and electromyogram (EMG) data of the subjects were collected simultaneously. Use a 64-channel wireless EEG system (NeuSen.W64, Neuracle, China) to collect EEG data, and its sampling frequency is 1000 HZ. Referring to the international 10 - 20 system, 39 channels were selected from 64 EEG channels for measurement. Before data collection, the impedance was kept below 5 kΩ by injecting conductive gel. Before the start of the whole experiment, collect the EEG data of each subject in the static state and actual kicking movement.

[0089] Step 2: Preprocess the EEG signals obtained in Step 1, including downsampling, denoising, filtering, and rereferencing.

[0090] Step 3: Calculate the power spectral density of the preprocessed EEG signals, and represent the power spectral density of each electrode in the form of a matrix:

[0091]

[0092] where, P i represents the average value of the power spectral density corresponding to the i-th electrode.

[0093] Step 4: Calculate the PDC connection matrix of the preprocessed EEG signals.

[0094] Step 5: Integrate the obtained P matrix and PDC matrix to obtain PPDC to measure the changes in the cerebral cortex of the subject during the static state, observation condition, illusion condition, and actual kicking movement. Estimate whether the subject has a sense of body ownership during the illusory movement based on the PPDC value.

[0095] As Figure 2 shown, the specific implementation of the illusion condition and the observation condition is as follows:

[0096] The subject sits on a chair with their legs relaxed and hanging naturally, and their arms in a relaxed state, maintaining a resting posture during the experiment. The illusion condition includes videos of the subject performing sitting front kicks with the left and right legs at leg-lifting angles of 30 degrees and 60 degrees respectively. To give the subject a more realistic illusion, we recorded videos of the subject performing sitting front kicks with the left and right legs at leg-lifting angles of 30 degrees and 60 degrees respectively one week before the experiment. As Figure 2 shown, we imported the recorded leg movement videos into a VR (virtual reality headset), adjusted their positions to create the illusion that the subject's own calves were moving (illusion condition). The observation condition used the same movement videos as the hallucination condition. The video was placed directly in front of the subject to observe both the video and their own legs simultaneously.

[0097] As Figure 3 shown, the processed EEG data was filtered into seven frequency bands (delta: 0.5 - 3.5 Hz, theta: 4 - 7.5 Hz, aleph: 8 - 11.5 Hz, beta1: 12 - 15.5 Hz, beta2: 16 - 19.5 Hz, gram1: 20 - 33.5 Hz, gram2: 34 - 98.5 Hz). For each subject and experimental action, we calculated the power spectral density from the filtered time series. Finally, the PSD matrix was averaged and summarized among the subjects. For these seven frequency bands, the PSD during the static state showed significant differences between the left and right legs of the subject and other conditions; there were not large significant differences in the PSD of the five frequency bands of δ, θ, α, β1, and β2 between the observation condition and the illusion condition; while in the γ (γ1 and γ2) frequency bands, there were significant differences between any two of the seven movement conditions (p < 0.01). In contrast, the γ2 frequency band was the most significant (p < 0.001), which may be because the γ2 frequency band is usually used as a binding tool for sensory processing of new information. Therefore, we focused on the γ2 frequency band for further investigation. In the γ2 frequency band, the PSD values of the left and right legs increased successively with the static state, observation condition, illusion condition, and control experiment; at the same time, the larger the leg-lifting angle, the larger the PSD value.

[0098] As Figure 4As shown, the PDC connection matrices of the EEG signals of the left and right legs under the observation conditions of the static state, the leg - raising angle of 30 degrees, the leg - raising angle of 60 degrees, the illusion conditions of the leg - raising angle of 30 degrees, the illusion conditions of the leg - raising angle of 60 degrees, and the control experiments of the leg - raising angles of 30 degrees and 60 degrees are presented. The PDC connection matrices are shown in the form of heatmaps. By comparing the heatmaps of different experimental conditions for the same - side leg, certain differences are found. At the same leg - raising angle, the values of the PDC connection matrix under the observation condition are larger than those under the static state, and smaller than those under the illusion condition; the PDC values under the illusion condition are smaller than the PDC values of the control experiment. That is to say, compared with the observation condition, the information flow between the various channels of the brain is more active under the illusion condition. We also found that the larger the leg - raising angle, the larger the PDC values corresponding to the observation condition, the illusion condition, and the control experiment.

[0099] As Figure 5 shown, we show the average values of C and Eglobal of the PDC connection matrices of the left and right legs under seven different experimental conditions at a network density of 15%. We observe that the average values of C and Eglobal under the illusion condition are larger, while the average values of C and Eglobal under the observation condition are smaller, indicating that the visual - motor illusion improves the correlation of EEG channels to a certain extent. At the same time, we also found that the average values of C and Eglobal of the illusion condition with a leg - raising angle of 60 degrees are greater than those of the illusion condition with a leg - raising angle of 30 degrees, suggesting that a higher - intensity visual - motor illusion has a greater effect on improving the correlation of EEG channels. For the left and right legs, the average values of C and Eglobal of the seven experimental conditions of the right leg are greater than those of the left leg, which may be because the subjects in our experiment are all right - leg - dominant, and the brain may respond more strongly to the right leg.

[0100] As Figure 6 shown, we show the PPDC of the γ2 frequency band of the left and right legs of the subjects under seven different experimental conditions. It can be seen from the figure that the PPDC values of the subjects under the illusion condition are generally greater than those under the observation condition. The contrast is more significant in the right leg, which may be because the visual - motor illusion of the right leg is more likely to activate the brain. We also found that when conducting illusion conditions with different leg - raising angles for the left and right legs, the values of PPDC in the frontal and central regions are larger.

[0101] Table 1 Comparison of the recognition performance of PPDC, PSD, and PDC

[0102]

[0103] ACC refers to the classification accuracy

[0104] Table 1 shows the classification accuracies of two groups of experiments. It can be seen that the accuracies of the PPDC index proposed by us are 81.1% and 87.3% respectively; these values are higher than the corresponding values measured by using PSD and PDC alone.

[0105] The embodiments of the present invention have been described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, without departing from the principle and spirit of the present invention, various changes, modifications, substitutions, and variations made to these embodiments still fall within the protection scope of the present invention.

Claims

1. A method for identifying ownership awareness of different moving bodies based on EEG signals, characterized by: The following steps are involved: Step 1: Before the experiment begins, collect the EEG signals of the subjects in a static state and during the actual leg-lifting exercise; Step 2: collecting the EEG signals of the subjects when they are in the observation condition and the illusion condition; Step 3: preprocessing the EEG signals obtained in steps 1 and 2; Step 4: Calculate the power spectral density of the EEG signal after preprocessing in step 3; Step 5: Calculate the partial directed correlation values of the EEG signal after preprocessing in step 3 to obtain the PDC connection matrix; Step 6: The power spectral density obtained in step 4 is fused with the PDC connectivity matrix obtained in step 5 to obtain a new indicator PPDC, which is used to estimate whether the subject has a sense of body ownership when experiencing the visual motion illusion; Step 7: Input the PPDC matrix obtained in step 6 into the convolutional neural network to identify different motion conditions and verify the feasibility of PPDC; The step five specifically includes: The EEG signal is modeled using a multivariate autoregressive model. The order of the model is determined using the AIC criterion. The directed information flow from channel j to channel i at frequency f, i.e., the PDC value, is solved using the following formula: Then the representation of the PDC matrix is: Among them, pdc ij Refers to the PDC value between the i-th electrode and the j-th electrode, and the matrix is a 39×39 two-dimensional square matrix; The PPDC in step 6 is obtained by feature fusion based on the P matrix and the PDC connection matrix calculated in steps 4 and 5:

2. The method for distinguishing ownership awareness of different moving bodies based on EEG signals according to claim 1, characterized in that: The static state in step 1 is a state in which the subject sits on a chair with both legs naturally falling down without any movement; the actual leg-lifting movement is a state in which the subject's left and right legs perform normal forward kicking in a sitting position.

3. The method for distinguishing ownership awareness of different moving bodies based on EEG signals according to claim 1, characterized in that: In the steps, a 64-channel wireless EEG system is used to collect EEG data with a sampling frequency of 1000 Hz. 39 channels are selected from the 64 EEG channels for measurement. Before data collection, the impedance is maintained below 5 kΩ by injecting conductive glue. The illusion conditions include videos of the left and right legs performing a seated front kick with leg lift angles of 30 degrees and 60 degrees, respectively.

4. The method for distinguishing ownership awareness of different moving bodies based on EEG signals according to claim 1, characterized in that: The step three specifically includes: Step 3.1: Locate the channel according to the standard; Step 3.2: Perform channel selection: Select 39 channels from the 64 EEG channels for measurement. Step 3.3: Reduce the sampling rate: Reduce the sampling rate of the EEG signal to 250Hz; Step 3.4: Cleanline was used to remove sinusoidal artifacts in the scalp channel that were not effectively removed by the notch filter, and other artifacts were removed by independent component analysis. Step 3.5: Filtering: Use the FIR digital filter in EEGLAB to bandpass filter the denoised EEG signal and divide it into seven frequency bands (delta: 0.5-3.5Hz, theta: 4-7.5Hz, aleph: 8-11.5Hz, beta1: 12-15.5Hz, beta2: 16-19.5Hz, gram1: 20-33.5Hz, gram2: 34-98.5Hz); Step 3.6: Rereference the signal.

5. The method for distinguishing ownership awareness of different moving bodies based on EEG signals according to claim 1, characterized in that: The step 4 specifically includes: Step 4.1: Calculate the power spectral density of the seven frequency bands (delta: 0.5-3.5Hz, theta: 4-7.5Hz, aleph: 8-11.5Hz, beta1: 12-15.5Hz, beta2: 16-19.5Hz, gram1: 20-33.5Hz, gram2: 34-98.5Hz) for the subjects in the static state, the observation condition of the leg lift angle of 30 degrees, the observation condition of the leg lift angle of 60 degrees, the illusion condition of the leg lift angle of 30 degrees, the illusion condition of the leg lift angle of 60 degrees, and the actual movement of the leg lift angles of 30 degrees and 60 degrees; Step 4.2: Average the power spectral density values within each frequency range and express the power spectral density of each electrode in matrix form: Among them, P i Represents the average value of the power spectral density corresponding to the i-th electrode.

Images