Time optimization method based on power spectrum density heat map

Abstract

The present invention provides a time optimization method based on a power spectrum density heat map, comprising: sliding a time window in a whole section of an electroencephalogram signal on a C3 or C4 channel using a specified time window length W and a time interval Δt to obtain a plurality of time windows, and calculating the power spectrum density of the electroencephalogram signal in each time window; drawing a power spectrum density heat map based on the power spectrum density, and smoothing the power spectrum density heat map using an isotropic heat diffusion equation; performing a first-order difference calculation on the smoothed power spectrum density heat map in a time direction, and then performing a square sum calculation on a target frequency range in a frequency direction to obtain a change curve; searching for a time t* in the change curve, sliding the time window at a time interval Δt in a training set data interval [t*-W / 2, t*+W / 2] to obtain (W / Δt) training sets; performing classification calculation on the training sets, and obtaining an optimal time point according to the calculation results.

Description

Technical Field

[0001] The present invention relates to the field of brain-computer interface technology, and in particular to a time optimization method based on a power spectral density heat map. Background Art

[0002] Accurately classifying the binary classification task of EEG signals is the core task of the brain-computer interface system based on motor imagery EEG signals. When performing motor imagery tasks, the EEG signals will show ERD / ERS phenomenon in specific frequency bands (motor rhythm) (the α band (8-12Hz) and β band (13-30Hz) of the contralateral sensorimotor cortex EEG signals show a decrease in amplitude, called ERD, which represents a decrease in the amplitude of the activated cortical EEG signals; the α band (8-12Hz) and β band (13-30Hz) of the cerebral cortex signals on the ipsilateral side of the moving hand show an increase in amplitude, called ERS, which represents an increase in the amplitude of the corresponding cortex in a resting state). Currently, most existing technologies classify motor imagery EEG signals within the task time through machine learning, and the EEG signals used are often signals within the entire task time.

[0003] Due to the unique complexity of EEG signals based on motor imagery, even if there is a trigger mechanism in the experimental paradigm, it is difficult to determine whether the subject is performing the imagination task or when he starts to perform the imagination task. Therefore, the EEG signals used for classification often have a low signal-to-noise ratio, and the signal-to-noise ratio of EEG signals performing the same motor imagery task in different subjects or the same subject in different states often has a large difference; traditional machine learning algorithms use EEG signals during the entire task time. When the signal-to-noise ratio of the signal used is high, a more ideal classification effect can be achieved, but when the signal-to-noise ratio of the signal used is low, the classification results often cannot meet actual usage needs. Summary of the invention

[0004] The purpose of the present invention is to provide a time optimization method based on power spectral density heat map, which can solve the problem that it is difficult to determine whether the subject is performing the imagination task or when he starts to perform the imagination task using EEG signals during the entire motor imagery task time, resulting in low classification accuracy.

[0005] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0006] A time optimization method based on a power spectrum density heat map comprises the following steps:

[0007] Step 1: Slide the time window in the entire EEG signal on the C3 and C4 channels using the specified time window length W and time interval Δt to obtain multiple time windows, and calculate the power spectrum density of the EEG signal in each time window;

[0008] Step 2: based on the power spectrum density, draw a power spectrum density heat map, and use the isotropic heat diffusion equation to smooth the power spectrum density heat map;

[0009] Step 3: perform a first-order difference calculation on the smoothed power spectrum density heat map in the time direction, and then perform a square sum calculation on the target frequency range in the frequency direction to obtain a change curve;

[0010] Step 4: Find the time t* in the change curve, and slide the time window at a time interval Δt in the training set data [t*-W / 2, t*+W / 2] to obtain (W / Δt) training sets;

[0011] Step 5: perform classification calculation on the training set and obtain the optimal time point according to the calculation result.

[0012] Preferably, in step 1, the power spectrum density of the EEG signal in each time window is calculated, and the calculation formula is as follows:

[0013]

[0014]

[0015] Among them, P is the signal power, E is the signal energy, x(t) is the time domain signal, x(f) is the frequency domain signal, T is the signal period, and x is the signal.

[0016] Preferably, in step 2, drawing a power spectrum density heat map includes:

[0017] The power spectrum density of the EEG signal in each time window is calculated as a column vector;

[0018] By sliding the time window, multiple column vectors are obtained and formed into a matrix;

[0019] The matrix is ​​visualized to obtain the power spectrum density heat map, whose ordinate is frequency and whose abscissa is time.

[0020] Preferably, in step 2, the calculation formula for smoothing is as follows:

[0021]

[0022] Among them, μ is the power spectrum density heat map, the partial derivative of x is the change in the time direction, the partial derivative of y is the change in the frequency direction, and the partial derivative of t is the change at the next moment in each iteration. After a certain number of iterations, the smoothing of the power spectrum density heat map μ is completed, and D is the diffusion coefficient.

[0023] Preferably, the step 3 specifically includes:

[0024] The first-order difference calculation of the smoothed power spectrum density heat map in the time direction is performed, and the calculation formula is as follows:

[0025] Δu t =u t+Δt –u t , where Δt represents the time interval, u t Indicates the power value at each frequency point at time t;

[0026] Then, the square sum of the target frequency range is calculated in the frequency direction. The calculation formula is as follows:

[0027] S=k1 2 +k2 2 +…+k n 2 , where k is the first-order difference change value obtained by differential calculation at each frequency point;

[0028] The calculation results are processed in a data visualization manner to obtain a change curve, wherein the ordinate of the change curve is the degree of change of each frequency, and the abscissa is the time point, thereby reflecting the degree of change of the energy of the EEG signal at a specific time point.

[0029] Preferably, in step 4, finding the time t* includes:

[0030] Obtaining a first-order derivative of a time vector on the variation curve;

[0031] According to the calculation results, the midpoints of the maximum peak and the minimum peak are solved to obtain the time t*.

[0032] Preferably, in step 5, obtaining the optimal time point includes:

[0033] Combine feature extraction and pattern recognition with machine learning methods to perform classification calculations on the training set;

[0034] Compare the classification calculation results, and select the signal time starting point corresponding to the training set with the highest accuracy as the optimal time point, which is the final preferred time.

[0035] Compared with the prior art, the beneficial effect achieved by the present invention is that: the present invention calculates the difference in the time direction through the power spectral density heat map to reflect the energy change, calculates the square sum of the energy change at each frequency to measure the degree of drastic change, and combines the first-order derivative to obtain the time t*, thereby determining the time interval in which the EEG signal shows a specific regularity, and then combines the classification calculation to obtain the optimal time point, and then confirms the starting time interval for the subject to perform the motor imagery task, so as to improve the classification accuracy of the EEG signal in the motor imagery task.

[0036] It should be understood that all combinations of the aforementioned concepts and the additional concepts described in more detail below can be considered as part of the inventive subject matter of the present disclosure as long as such concepts are not mutually inconsistent. In addition, all combinations of the claimed subject matter are considered as part of the inventive subject matter of the present disclosure.

[0037] The foregoing and other aspects, embodiments and features of the present invention can be more fully understood from the following description in conjunction with the accompanying drawings. Other additional aspects of the present invention, such as the features and / or beneficial effects of the exemplary embodiments, will be apparent from the following description or learned from the practice of the specific embodiments according to the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings are not intended to be drawn to scale. In the accompanying drawings, each identical or nearly identical component shown in each figure may be represented by the same reference numeral. For clarity, not every component is labeled in each figure. Embodiments of various aspects of the present invention will now be described by way of example and with reference to the accompanying drawings.

[0039] Figure 1 It is a schematic flow chart of a time optimization method based on a power spectrum density heat map provided in one embodiment of the present invention;

[0040] Figure 2 is a schematic diagram of a public data set experimental paradigm selected in one embodiment of the present invention;

[0041] Figure 3 is a power spectral density heat map formed by a public data set experiment provided in one embodiment of the present invention;

[0042] Figure 4 is a power spectral density heat map after smoothing provided in one embodiment of the present invention;

[0043] Figure 5 It is a schematic diagram of the gradient change of the power spectrum density heat map in the time direction provided in one embodiment of the present invention;

[0044] Figure 6 It is a schematic diagram of the time point t* at which the intensity of the power spectrum density heat map decreases the most after solving the first-order derivative provided in one embodiment of the present invention;

[0045] Figure 7 It is a schematic diagram comparing the fixed time starting point used by a traditional brain-computer interface-based rehabilitation system provided in one embodiment of the present invention and the signal time starting point optimized by the method of the present invention. DETAILED DESCRIPTION

[0046] In order to better understand the technical content of the present invention, specific embodiments are given and described as follows in conjunction with the accompanying drawings.

[0047] Various aspects of the present invention are described in this disclosure with reference to the accompanying drawings, in which many illustrative embodiments are shown. The embodiments of the present disclosure are not necessarily intended to include all aspects of the present invention. It should be understood that the various concepts and embodiments introduced above, as well as those described in more detail below, can be implemented in any of many ways, because the concepts and embodiments disclosed by the present invention are not limited to any implementation. In addition, some aspects disclosed by the present invention can be used alone or in any appropriate combination with other aspects disclosed by the present invention.

[0048] According to an embodiment of the present invention, a time optimization method based on a power spectral density heat map includes:

[0049] Step 1: Slide the time window in the entire EEG signal on the C3 and C4 channels using the specified time window length W and time interval Δt to obtain multiple time windows, and calculate the power spectrum density of the EEG signal in each time window;

[0050] Step 2: Based on the power spectrum density, draw a power spectrum density heat map, and use the isotropic heat diffusion equation to smooth the power spectrum density heat map;

[0051] Step 3: Perform a first-order difference calculation on the smoothed power spectrum density heat map in the time direction, and then perform a square sum calculation on the target frequency range in the frequency direction to obtain a change curve;

[0052] Step 4: Find the time t* in the change curve, and slide the time window at a time interval Δt in the training set data interval [t*-W / 2, t*+W / 2] to obtain (W / Δt) training sets;

[0053] Step 5: Perform classification calculation on the training set and obtain the optimal time point based on the calculation results.

[0054] Combine the following Figure 1 The flowcharts shown and some preferred or optional examples of the present invention more specifically describe the implementation and / or effects of certain examples of the present invention.

[0055] Sliding time window

[0056] The EEG cap was used to obtain the EEG signals of the subjects when they performed the motor imagery task;

[0057] The EEG signals generated from the beginning of preparation to the completion of the motor imagery task on the C3 and C4 channels (the C3 and C4 channels are located in the motor function area of ​​the brain and are the most active areas for left and right hand motor imagery tasks) are used as the whole EEG signals;

[0058] Define the time window length as W and the time interval as Δt;

[0059] The time window length W and the time interval Δt are used to slide the time window in the entire EEG signal on the C3 and C4 channels to obtain multiple time windows.

[0060] In this step, the method of using the EEG cap to obtain EEG signals and slide the time window can be implemented by means of the prior art.

[0061] Calculate power spectral density

[0062] In the above step 1, the power spectrum density of the EEG signal in each time window is calculated, and the calculation formula is as follows:

[0063]

[0064]

[0065] Among them, P is the signal power, E is the signal energy, x(t) is the time domain signal, x(f) is the frequency domain signal, T is the signal period, and x is the signal.

[0066] It should be noted that the power spectral density is the signal power within a unit frequency band, which represents the distribution of the signal power in the frequency domain. This embodiment calculates the power spectral density of the EEG signal in each time window to reflect the energy of the EEG signal vibration within each frequency range, and obtains the regular vibration and irregular vibration of the EEG signal, so as to find the time position of the regular vibration presented by the EEG signal when performing the motor imagery task, so as to improve the efficiency and accuracy of the subsequent step of obtaining the time t*.

[0067]

Drawing power spectrum density heat map

[0068] In the above step 2, drawing a power spectrum density heat map based on the power spectrum density includes:

[0069] The power spectrum density of the EEG signal in each time window is calculated as a column vector;

[0070] By sliding the time window, multiple column vectors are obtained and formed into a matrix;

[0071] The matrix is ​​visualized to obtain a power spectrum density heat map, with the ordinate being frequency and the abscissa being time.

[0072]

Smoothing

[0073] In the above step 2, the power spectrum density heat map is smoothed using the isotropic heat diffusion equation. The smoothing calculation is as follows:

[0074]

[0075] Among them, μ is the power spectrum density heat map, the partial derivative of x is the change in the time direction, the partial derivative of y is the change in the frequency direction, and the partial derivative of t is the change at the next moment in each iteration. After a certain number of iterations, the smoothing of the power spectrum density heat map μ is completed, and D is the diffusion coefficient.

[0076] Preferably, this embodiment smoothes the power spectral density heat map through the isotropic heat diffusion equation, uses the Laplace operator (the Laplace operator is a second-order differential operator in Euclidean space and is isotropic) to perform low-pass filtering and denoising, and smoothes the high-frequency components to enhance the diffusion direction of the power spectral density heat map (the direction of propagation with real drastic changes), so as to better find more obvious time boundaries from the power spectral density heat map, and improve the accuracy of the subsequent acquisition of the change curve to reflect the accuracy of the energy change of the EEG signal at a specific time point.

[0077]

Obtain change curve

[0078] The aforementioned step 3 specifically includes:

[0079] The first-order difference calculation of the smoothed power spectrum density heat map in the time direction is performed, and the calculation formula is as follows:

[0080] Δu t =u t+Δt –u t , where Δt represents the time interval, u t Indicates the power value at each frequency point at time t;

[0081] Then, the square sum of the target frequency range is calculated in the frequency direction. The calculation formula is as follows:

[0082] S=k1 2 +k2 2 +…+k n 2 , where k is the first-order difference change value obtained by differential calculation at each frequency point;

[0083] The calculation results are processed in a data visualization manner to obtain a change curve. The ordinate of the change curve is the degree of change of each frequency, and the abscissa is the time point, which reflects the degree of change of the energy of the EEG signal at a specific time point.

[0084]

Get time t*

[0085] In the aforementioned step 4, searching for the time t* in the change curve includes:

[0086] Obtain the first-order derivative of the time vector on the changing curve;

[0087] According to the calculation results, the midpoint of the maximum peak and the minimum peak is solved to obtain the time t*;

[0088] The characteristic of time t* is that after a large change in the EEG signal (the intensity of the power spectral density heat map decreases the most after solving the first-order derivative), there is a continuous stable interval (regular vibration) to determine the specific regularity reflected after performing the motor imagery task.

[0089] In this step, the first-order derivative can be achieved by means of the prior art.

[0090] [Get training set]

[0091] Collect the subject data in the brain-computer interface-based rehabilitation system to form training set data;

[0092] In the training set data interval [t*-W / 2, t*+W / 2], the time window is slid at a time interval of Δt to obtain (W / Δt) training sets.

[0093] Furthermore, taking the 0.2s interval as an example, starting points are selected at intervals of t*-0.5, t*-0.3, t*-0.1, t*+0.1, t*+0.3, t*+0.5 in the training set data, and EEG signals with a time length of 2s are extracted to form 10 (2S / 0.2S) training sets with a time length of 2S.

[0094] It should be noted that the training set acquisition method provided in this embodiment is to select the time point that best reflects the regular part of the EEG signal when executing the motor imagination task from the EEG signal during the entire period of running the imagination task, and provide a reference benchmark for the subsequent EEG signal acquisition through machine learning training, so that the EEG signal acquisition data corresponding to the subsequent selected time point is more accurate.

[0095]

Get the preferred time point

[0096] In the above step 5, the training set is classified and calculated, and the optimal time point is obtained according to the calculation results, including:

[0097] Combining the machine learning method of feature extraction and pattern recognition, this embodiment uses the CSP feature extraction algorithm to extract the spatial features of the training set, and combines the SVM algorithm for pattern recognition to complete the classification calculation;

[0098] Compare the classification calculation results, and select the signal time starting point corresponding to the training set with the highest accuracy as the optimal time point, which is the final preferred time.

[0099] Still taking the 0.2s interval as an example, starting points are selected at intervals of t*-0.5, t*-0.3, t*-0.1, t*+0.1, t*+0.3, t*+0.5 in the training set data, and EEG signals with a time length of 2s are extracted to form 10 (2S / 0.2S) training sets with a time length of 2S. After classification calculation, the training set corresponding to t*-0.1 as the signal time starting point has the highest accuracy, so t*-0.1 is selected as the optimal time point.

[0100] Preferably, this embodiment takes into account the differences in the regularities of EEG signals of different subjects. Therefore, the time optimization method based on the power spectral density heat map provided in this embodiment obtains the change curve to find the specific regular time t* that can be determined for executing the motor imagery task, and uses the training set to perform classification calculations (training classifiers) to find the signal time point with the highest accuracy as the preferred time point (the time point when the subject starts to perform the imagination task).

[0101] Preferably, taking t*-0.1 as the optimal time point in the above example, when applied to other EEG signal classifications, the method of the present invention can be used to obtain a new specific regular time t* reflected in determining the execution of the motor imagery task. + (i.e., the time t* obtained in the method of the present invention), and the new time t* + -0.1 is taken as the starting point (the time point when the subjects start to perform the imagination task), and the EEG signal with a time length of 2S is extracted, and then the subsequent EEG signal feature extraction and classification work is carried out, which further improves the classification accuracy of the EEG signal.

[0102] The following is combined with Figure 2-7 , further illustrating the specific implementation of the above steps.

[0103] This embodiment uses the public data set BCI Competition III: Data Set IIIa (a neural network motor imagery data set) to verify the real effect of the method of the present invention. Three subjects perform four motor imagery tasks (left hand, right hand, foot and tongue), and their 60-channel EEG signals are collected with a sampling frequency of 250 Hz.

[0104] Combination Figure 2In a single experiment, the subjects sat in an armchair and performed a randomly assigned motor imagery task according to the prompts. They remained silent for the first 2 seconds. At t=2s, a prompt sounded and a fixed cross appeared on the screen. Starting from t=3s, an arrow pointed up, down, left or right and lasted for 1s. While the arrow was displayed, the subjects were asked to perform the corresponding motor imagery task until the cross sign disappeared at t=7s, ending the imagination.

[0105] The experiment was divided into a verification group and a comparison group, and the left-hand and right-hand motor imagery EEG signals were classified respectively. The verification group used TS+CSP+SVM (adopting the time optimization method of the present invention) to classify the EEG signals, and the comparison group used CSP+SVM (traditional classification method) to classify the EEG signals for the fixed 3-7s signals.

[0106] The specific steps are as follows:

[0107]

Data preprocessing

[0108] Set the time window length to 2s and the sliding interval to 10 / sampling frequency (the sampling frequency here is 250Hz).

[0109] The EEG signal data of three subjects were collected when they performed four motor imagery tasks (left hand, right hand, foot and tongue). The signals classified as left hand and right hand in the EEG signal data were segmented, and the alpha and beta bands (5-30Hz) were extracted.

[0110] Extract the entire EEG signal on C3 and C4 channels from the 60-channel signal.

[0111]

Drawing power spectrum density heat map

[0112] Combination Figure 3 ,The entire EEG signals extracted on C3 and C4 channels are divided into signal time ,sliding interval (i.e. 7 / (10 / 250)) segments by sliding time window, and the power ,spectral density difference of each segment is calculated respectively, and the calculated results are plotted as a ,success rate spectral density heat map.

[0113]

Smoothing

[0114] Combination Figure 4 , the power spectral density heat map is smoothed to reduce image noise, so as to better find more obvious time boundaries from the heat map and obtain a smoothed heat map.

[0115] [Get waveform curve]

[0116] Combination Figure 5, calculate the square of the gradient of the smoothed heat map in the time direction, sum it in the alpha and beta bands respectively, and form a waveform curve (i.e., the change curve provided in the above steps of this embodiment), where, Figure 5 The solid line in represents the beta band, and the dotted line represents the alpha band. The obtained waveform curve is smoothed using a sliding average, and the waveform curve is used to measure the intensity of change of the heat map in the time direction.

[0117]

Get time t*

[0118] Combination Figure 6 By calculating the first-order derivative of the above waveform curve, the time point t* at which the intensity of the change of the thermal diagram in the time direction decreases the most (that is, the time point t* on the change curve to be obtained in this embodiment) is obtained. According to Figure 6 The time point corresponding to the minimum value is the time point t* when the intensity of the heat map changes decreases the most.

[0119] [Get the time point when the subject starts to perform the imagination task]

[0120] In the range of 1s before and after time t*, the time window was slid with a time interval of 0.2 seconds to extract EEG signals with a time length of 2s, forming 10 new training sets with a time length of 2S (2S / 0.2S). The 10 new training sets were classified using the common spatial pattern (CSP) and support vector machine (SVM) respectively, and the time point corresponding to the result with the highest classification accuracy was selected as the time point when the subjects started to perform the imagination task.

[0121] Combination Figure 7 , the solid line in the left hand and the dotted line in the right hand are the fixed starting points of the traditional classification method, the dotted line in the left hand and the solid line in the right hand are the signal starting points (i.e., the time point when the subject starts to perform the imagination task) selected by the method of the present invention, according to Figure 7 It can be seen intuitively from the diagram that the position of the signal starting point optimized by the method of the present invention is completely different from the fixed starting point of the traditional classification method.

[0122]

Experimental results

[0123] The experimental results are shown in Table 1:

[0124] Table 1: Classification accuracy comparison data table.

[0125]

[0126] Combined with Table 1, according to the experimental comparison results, in the classification tasks with higher signal quality (the classification accuracy of the traditional classification method>70% available standard), the method of the present invention obtained a classification accuracy similar to that of the traditional classification method, and in the classification tasks with lower signal quality, the method of the present invention (TS+CSP+SVM) obtained a higher classification accuracy than the traditional classification method (i.e., the comparison group CSP+SVM).

[0127] It needs to be explained again that the traditional classification method is to classify the EEG signals during the entire motor imagery task time, but the signal-to-noise ratio of the EEG signals of different subjects or the same subject performing the same motor imagery task in different states often has a large difference. Therefore, the EEG signals used for classification often have a lower signal-to-noise ratio. Since the traditional classification method has a fixed trigger starting point, it cannot determine whether the subject is performing the imagination task or when he starts to perform the imagination task, resulting in unsatisfactory classification results. The time optimization method provided in this embodiment can find the signal starting point when the subject starts to perform the motor imagery task as a reference benchmark, thereby achieving a more ideal classification effect.

[0128] Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. A person with ordinary knowledge in the technical field to which the present invention belongs may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be determined by the definition of the claims.

Claims

1. A time optimization method based on power spectrum density heat map, characterized in that: include: Step 1: Slide the time window in the entire EEG signal on the C3 and C4 channels using the specified time window length W and time interval Δt to obtain multiple time windows, and calculate the power spectrum density of the EEG signal in each time window; Step 2: based on the power spectrum density, draw a power spectrum density heat map, and use the isotropic heat diffusion equation to smooth the power spectrum density heat map; Step 3: perform a first-order difference calculation on the smoothed power spectrum density heat map in the time direction, and then perform a square sum calculation on the target frequency range in the frequency direction to obtain a change curve; Step 4: Find the time t* in the change curve, and slide the time window at a time interval Δt in the training set data [t*-W / 2, t*+W / 2] to obtain (W / Δt) training sets; Find the time t* in the change curve, including: Obtaining a first-order derivative of a time vector on the variation curve; According to the calculation results, the midpoint of the maximum peak and the minimum peak is solved to obtain the time t*; Step 5: performing classification calculation on the training set and obtaining the optimal time point according to the calculation result, including: Combine feature extraction and pattern recognition with machine learning methods to perform classification calculations on the training set; By comparing the classification calculation results, the signal time starting point corresponding to the training set with the highest accuracy is selected as the optimal time point, which is the final preferred time. The preferred time is the time point when the subject starts to perform the imagination task.

2. The time optimization method based on power spectrum density heat map according to claim 1 is characterized in that: In the above step 1, the power spectrum density of the EEG signal in each time window is calculated, and the calculation formula is as follows: Among them, P is the signal power, E is the signal energy, x(t) is the time domain signal, x(f) is the frequency domain signal, T is the signal period, and x is the signal.

3. The time optimization method based on power spectrum density heat map according to claim 1 or 2, characterized in that: In the above step 2, drawing a power spectrum density heat map based on the power spectrum density includes: The power spectrum density of the EEG signal in each time window is calculated as a column vector; By sliding the time window, multiple column vectors are obtained and formed into a matrix; The matrix is ​​visualized to obtain the power spectrum density heat map, where the ordinate is frequency and the abscissa is time.

4. The time optimization method based on power spectrum density heat map according to claim 1, characterized in that: In the above step 2, the power spectrum density thermogram is smoothed using the isotropic heat diffusion equation, and the smoothing calculation is as follows: Among them, u is the power spectrum density heat map, the partial derivative of x is the change in the time direction, the partial derivative of y is the change in the frequency direction, and the partial derivative of t is the change in the next moment in each iteration. After a certain number of iterations, the smoothing of the power spectrum density heat map u is completed, and D is the diffusion coefficient.

5. The time optimization method based on power spectrum density heat map according to claim 1, characterized in that: In the above step 3, the first-order difference calculation is performed on the smoothed power spectrum density heat map in the time direction, and the calculation formula is as follows: Δu t =u t+Δt –u t , where Δt represents the time interval, u t Indicates the power value at each frequency point at time t; Then, the square sum of the target frequency range is calculated in the frequency direction. The calculation formula is as follows: S=k1 2 +k2 2 +…+k n 2 , where k is the first-order difference change value obtained by differential calculation at each frequency point; The calculation results are processed in a data visualization manner to obtain a change curve, wherein the ordinate of the change curve is the degree of change of each frequency, and the abscissa is the time point, thereby reflecting the degree of change of the energy of the EEG signal at a specific time point.

Images