WO2025118317A1 - 一种神经电脉冲检测方法、系统及终端 - Google Patents
一种神经电脉冲检测方法、系统及终端 Download PDFInfo
- Publication number
- WO2025118317A1 WO2025118317A1 PCT/CN2023/138053 CN2023138053W WO2025118317A1 WO 2025118317 A1 WO2025118317 A1 WO 2025118317A1 CN 2023138053 W CN2023138053 W CN 2023138053W WO 2025118317 A1 WO2025118317 A1 WO 2025118317A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- electrical pulse
- neural
- probability density
- neural electrical
- pulse signal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/369—Electroencephalography [EEG]
- A61B5/372—Analysis of electroencephalograms
- A61B5/374—Detecting the frequency distribution of signals, e.g. detecting delta, theta, alpha, beta or gamma waves
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/388—Nerve conduction study, e.g. detecting action potential of peripheral nerves
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Definitions
- the present application relates to the field of artificial intelligence technology, and in particular to a neural electrical pulse detection method, system, terminal and computer-readable storage medium.
- neural electrical pulse detection is a core topic in the field of neuroscience and biomedical engineering. It involves accurately and efficiently identifying the discharge activity of neurons from complex EEG signals. Technological advances, especially the development of high-throughput electrode arrays, have posed new challenges to this task, namely how to overcome the difficulty of quickly and accurately detecting sparse neural electrical pulses from hundreds or thousands of electrode channels.
- the present application provides a neural electrical pulse detection method, system, terminal and computer-readable storage medium to solve the problems of neural electrical pulse detection methods in the prior art, such as low noise resistance, poor specificity, loss of a large amount of waveform information, and complex template matching process.
- the present application proposes a method for detecting neural electrical pulses, the method comprising:
- the acquiring of the original neural electrophysiological signal, filtering and threshold detection of the original neural electrophysiological signal, and obtaining the first neural electrical pulse signal sample specifically includes:
- the first filtered electrical pulse signal is threshold-detected using a threshold detection parameter to obtain the first neural electrical pulse signal sample.
- making a skewed distribution assumption for the first neural electrical pulse signal sample, obtaining a first probability density parameter, a position parameter and a scale parameter of the skewed distribution of the first neural electrical pulse signal sample according to the moment estimation of the skewed distribution, and obtaining a first probability density function according to the first probability density parameter, the position parameter and the scale parameter specifically includes:
- the skewed distribution assumption is made for the first neural electrical pulse signal sample: Y ⁇ SN( ⁇ , ⁇ 2 , ⁇ );
- Y represents the first neural electrical pulse signal sample
- SN represents skewed distribution
- ⁇ R is the location parameter
- ⁇ R + is the scale parameter
- ⁇ is the first probability density parameter
- ⁇ 0 is a tilted variable
- X is a sample that obeys the standard skewed distribution
- ⁇ ( ⁇ x) Cumulative probability distribution function
- x is a random variable.
- the using a kernel density estimation method to calculate the second neural electrical pulse signal sample to obtain a second probability density function specifically includes:
- the calculation formula of the kernel density estimation method is:
- the kernel function K is a Gaussian kernel xi is the sample point, n is the number of sample points, h is the bandwidth, and the bandwidth is
- the using a kernel density estimation method to calculate the second neural electrical pulse signal sample to obtain a second probability density function specifically includes:
- the calculation formula of the kernel density estimation method is:
- the kernel function K is a linear kernel xi is the sample point, n is the number of sample points, h is the bandwidth, and the bandwidth is
- the using a kernel density estimation method to calculate the second neural electrical pulse signal sample to obtain a second probability density function specifically includes:
- the calculation formula of the kernel density estimation method is:
- the kernel function K is the cosine kernel
- ⁇ is the circumference of a circle
- xi is the sample point
- n is the number of sample points
- h is the bandwidth
- the bandwidth is the bandwidth
- the calculating the information divergence between the first probability density function and the second probability density function, judging whether the first neural electrical pulse signal sample belongs to a neural electrical pulse according to the information divergence, and outputting the judgment result specifically includes:
- KDE(i) represents the second probability density function
- KL(SN,KDE) represents the information divergence
- the judgment result is that the first neural electrical pulse signal sample belongs to a neural electrical pulse.
- the present application also proposes a neural electrical pulse detection system, the neural electrical pulse detection system comprising:
- a signal preprocessing module used for acquiring an original neural electrophysiological signal, performing filtering and threshold detection on the original neural electrophysiological signal, and obtaining a first neural electrical pulse signal sample;
- a signal moment estimation module configured to make a skewed distribution assumption for the first neural electrical pulse signal sample, obtain a first probability density parameter, a position parameter and a scale parameter of the skewed distribution of the first neural electrical pulse signal sample according to the moment estimation of the skewed distribution, and obtain a first probability density function according to the first probability density parameter, the position parameter and the scale parameter;
- a parameter determination module configured to determine whether the first probability density parameter is less than zero, and if so, use the first neural electrical pulse signal sample as a second neural electrical pulse signal sample;
- a signal density estimation module used to calculate the second neural electrical pulse signal sample using a kernel density estimation method to obtain a second probability density function
- a signal determination module used to calculate the information between the first probability density function and the second probability density function Divergence, judging whether the first neural electrical pulse signal sample belongs to a neural electrical pulse according to the information divergence, and outputting a judgment result.
- the present application also proposes a terminal, which includes: a memory, a processor, and a neural electrical pulse detection program stored in the memory and executable on the processor, wherein the neural electrical pulse detection program implements the steps of the neural electrical pulse detection method as described above when executed by the processor.
- the present application also proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a neural electrical pulse detection program, and when the neural electrical pulse detection program is executed by a processor, the steps of the neural electrical pulse detection method as described above are implemented.
- the present application obtains the original neural electrophysiological signal, performs filtering and threshold detection on the original neural electrophysiological signal, thereby reducing the noise of the original neural electrophysiological signal, and obtains the first neural electric pulse signal sample, thereby improving the anti-noise ability; secondly, the present application makes a skewed distribution assumption on the first neural electric pulse signal sample, obtains the first probability density parameter, position parameter and scale parameter of the skewed distribution of the first neural electric pulse signal sample according to the moment estimation of the skewed distribution, and obtains the first probability density function according to the first probability density parameter, position parameter and scale parameter, so as to facilitate the use of density estimation algorithm and judgment of whether the first neural electric pulse signal sample belongs to a neural electric pulse in subsequent steps; thirdly, The present application determines whether the first probability density parameter is less than zero.
- the first neural electric pulse signal sample is used as the second neural electric pulse signal sample to avoid the first probability density parameter that does not meet the conditions from entering the next step; in addition, the present application uses the kernel density estimation method to calculate the second neural electric pulse signal sample to obtain the second probability density function, which can quickly and efficiently perform neural electric pulse detection; in addition, the present application calculates the information divergence between the first probability density function and the second probability density function, determines whether the first neural electric pulse signal sample belongs to a neural electric pulse according to the information divergence, and outputs the judgment result, which has good specificity, does not lose a lot of waveform information, and avoids the complex process of template matching.
- FIG1 is a flow chart of a preferred embodiment of the neural electrical pulse detection method of the present application.
- FIG2 is a diagram of the original neural electrophysiological signal in the neural electrical pulse detection method of the present application.
- FIG3 is a sample diagram of the first neural electrical pulse signal in the neural electrical pulse detection method of the present application.
- FIG4 is a first type of a large number of neural electrical pulse signal samples in the neural electrical pulse detection method of the present application
- FIG5 is a sample diagram of the second type of first neural electrical pulse signal in the neural electrical pulse detection method of the present application.
- FIG6 is a statistical histogram of the first type of noise samples in the neural electrical pulse detection method of the present application.
- FIG7 is a statistical histogram of the second type of noise samples in the neural electrical pulse detection method of the present application.
- FIG8 is a statistical histogram of the third type of noise samples in the neural electrical pulse detection method of the present application.
- FIG9 is a statistical histogram of the fourth type of noise samples in the neural electrical pulse detection method of the present application.
- FIG10 is a histogram of a single neural electrical pulse signal in the neural electrical pulse detection method of the present application.
- FIG11 is a sample theoretical SN probability density distribution diagram in the neural electrical pulse detection method of the present application.
- FIG12 is a probability density distribution diagram after Gaussian kernel density estimation in the neural electrical pulse detection method of the present application.
- FIG13 is a schematic diagram showing the principle of a preferred embodiment of the neural electrical pulse detection system of the present application.
- FIG. 14 is a schematic diagram of the operating environment of a preferred embodiment of the terminal of the present application.
- the present application provides a neural electrical pulse detection method, system, terminal and computer-readable storage medium to solve the problems of neural electrical pulse detection methods in the prior art, such as low noise resistance, poor specificity, loss of a large amount of waveform information and complex template matching process.
- Figure 1 is a flow chart of a preferred embodiment of the neural electric pulse detection method of the present application
- Figure 2 is a diagram of the original neural electrophysiological signal in the neural electric pulse detection method of the present application
- Figure 3 is a diagram of the first neural electric pulse signal sample in the neural electric pulse detection method of the present application
- Figure 4 is a diagram of a first type of a large number of neural electric pulse signal samples in the neural electric pulse detection method of the present application
- Figure 5 is a diagram of a second type of first neural electric pulse signal sample in the neural electric pulse detection method of the present application
- Figure 6 is a statistical histogram of the first type of noise samples in the neural electric pulse detection method of the present application
- Figure 7 is a statistical histogram of the second type of noise samples in the neural electric pulse detection method of the present application
- Figure 8 is a statistical histogram of the third type of noise samples in the neural electric pulse detection method of the present application
- Figure 9 is a statistical histogram of the fourth type of noise samples in the neural electric pulse detection
- the present application proposes a method for detecting neural electrical pulses, wherein, as shown in FIG1 , the method for detecting neural electrical pulses comprises the following steps:
- Step S100 obtaining an original neural electrophysiological signal, filtering and threshold detection on the original neural electrophysiological signal, and obtaining a first neural electrical pulse signal sample.
- the original neural electrophysiological signal is obtained, the original neural electrophysiological signal is filtered and threshold detected, the original neural electrophysiological signal is denoised, and the first neural electrical pulse signal sample is obtained, thereby improving the anti-noise ability.
- An original neural electrophysiological signal is obtained, and the original neural electrophysiological signal is filtered using a sixth-order Butterworth filter to obtain a first filtered electrical pulse signal.
- the first cutoff frequency and the second cutoff frequency of the sixth-order Butterworth filter are 250 Hz and 7000 Hz respectively.
- the first filtered electrical pulse signal is threshold-detected using a threshold detection parameter to obtain the first neural electrical pulse signal sample.
- the threshold detection parameters are set so that the negative peak value of the signal is greater than 60 microvolts, and the peak-to-peak interval is set to be greater than 2 milliseconds, and the peak moment is taken as the base point, 1 millisecond forward and 1.5 milliseconds backward are regarded as the first neural electrical pulse signal sample.
- the collected signal is the original neural electrophysiological signal.
- the original neural electrophysiological signal is obtained.
- background noise and neural electrical pulse signals of concern to researchers are usually included.
- the original neural electrophysiological signal is filtered using a forward and backward sixth-order Butterworth filter to obtain a first filtered electrical pulse signal, wherein the first cutoff frequency and the second cutoff frequency of the sixth-order Butterworth filter are 250 Hz and 7000 Hz, respectively, the first cutoff frequency is the upper cutoff frequency, and the second cutoff frequency is the lower cutoff frequency.
- the threshold detection parameter is used to perform threshold detection on the first filtered electrical pulse signal to obtain a first neural electrical pulse signal sample.
- the threshold detection parameters are set to the negative peak value of the signal greater than 60 microvolts, and the peak-to-peak interval is set to be greater than 2 milliseconds, and the peak moment is taken as the base point, 1 millisecond is taken forward and 1.5 milliseconds is taken backward as the first neural electrical pulse signal sample, the threshold detection parameters are set to the negative peak value of the signal greater than 60 microvolts, and the peak-to-peak interval is greater than 2 milliseconds, as shown in Figure 3, and the peak moment is taken as the base point, 1 millisecond is taken forward and 1.5 milliseconds is taken backward as the potential first neural electrical pulse signal sample.
- Step S200 Make a skewed distribution assumption for the first neural electrical pulse signal sample, obtain a first probability density parameter, position parameter and scale parameter of the skewed distribution of the first neural electrical pulse signal sample according to the moment estimation of the skewed distribution, and obtain a first probability density function according to the first probability density parameter, the position parameter and the scale parameter.
- the distribution curve of the first neural electrical pulse signal sample belongs to a left-right asymmetric data frequency distribution, it is more in line with a skewed distribution. Therefore, a skewed distribution assumption is made for the first neural electrical pulse signal sample, and the first probability density parameter, position parameter and scale parameter of the skewed distribution of the first neural electrical pulse signal sample are obtained according to the moment estimation of the skewed distribution.
- the SN distribution assumption of neural electrical pulses in this application comes from the statistical inference of a large number of samples. As shown in Figures 4 and 5, they are statistical histograms of two types of neural electrical pulse signals, respectively. It can be seen that although the different types of neural electrical pulse signals are different in size, they all belong to the SN distribution, and their distribution parameter ⁇ is less than 0.
- the statistical histograms of noise samples of different categories show that the statistical histograms of noise samples of different categories have different performances and are not SN distributions with ⁇ 0.
- the step S200 making a skewed distribution assumption for the first neural electrical pulse signal sample, obtaining a first probability density parameter, a position parameter and a scale parameter of the skewed distribution of the first neural electrical pulse signal sample according to the moment estimation of the skewed distribution, and obtaining a first probability density function according to the first probability density parameter, the position parameter and the scale parameter, specifically includes:
- the skewed distribution assumption is made for the first neural electrical pulse signal sample: Y ⁇ SN( ⁇ , ⁇ 2 , ⁇ );
- Y represents the first neural electrical pulse signal sample
- SN represents skewed distribution
- ⁇ R is the location parameter
- ⁇ R + is the scale parameter
- ⁇ is the first probability density parameter
- ⁇ 0 is a tilted variable
- X is a sample that obeys the standard skewed distribution
- ⁇ ( ⁇ x) Cumulative probability distribution function
- x is a random variable.
- the Skew-Normal distribution (SN distribution) assumption is made for the first neural electrical pulse signal sample, and the formula can be obtained: Y ⁇ SN( ⁇ , ⁇ 2 , ⁇ ),
- Y represents the first neural electrical pulse signal sample
- SN represents skewed distribution
- ⁇ R is the location parameter
- ⁇ R + is the scale parameter
- ⁇ is the first probability density parameter
- ⁇ 0 is a tilted variable
- X is a sample that obeys the standard skewed distribution
- the mathematical formula of probability density function is: in, represents the standard Gaussian density function, ⁇ ( ⁇ x) is Cumulative probability distribution function, x is a random variable.
- the SN distribution probability density parameter of the first neural electrical pulse signal sample Y is estimated according to the moment of the skewed probability density distribution, thereby obtaining the theoretical probability density function of the first neural electrical pulse signal sample Y, namely, SN-pdf.
- the estimation diagrams of a single neural electrical pulse sample show that the estimation of a single neural electrical pulse sample and the use of Gaussian kernel density estimation are also in line with expectations.
- it is a histogram of a single neural electrical pulse signal, as shown in Figure 11, it is the theoretical SN probability density distribution of the sample, and as shown in Figure 12, it is the probability density distribution after Gaussian kernel density estimation.
- Step S300 determine whether the first probability density parameter is less than zero, and if so, use the first neural electrical pulse signal sample as the second neural electrical pulse signal sample.
- the first probability density parameter of the theoretical SN distribution is less than zero, that is, whether it satisfies the condition of ⁇ 0. If so, the second neural electrical pulse signal sample is obtained; if not, the first neural electrical pulse signal sample is rejected.
- Step S400 using a kernel density estimation method to calculate the second neural electrical pulse signal sample to obtain a second probability density function.
- the second neural electrical pulse signal sample is calculated using a kernel density estimation method to obtain a second probability density function, which can quickly and efficiently perform neural electrical pulse detection.
- the step S400: using a kernel density estimation method to calculate the second neural electrical pulse signal sample to obtain a second probability density function specifically includes:
- the calculation formula of the kernel density estimation method is:
- the kernel function K is a Gaussian kernel xi is the sample point, n is the number of sample points, h is the bandwidth, and the bandwidth is
- kernel density estimation is used to estimate the true probability density distribution (KDE-pdf) of the second neural electrical pulse signal sample.
- the Gaussian kernel density estimation method is used for calculation.
- the calculation formula of the Gaussian kernel density estimation method is: Among them, the kernel function is a Gaussian kernel xi is the sample point, n is the number of sample points, h is the bandwidth, and the bandwidth is The second probability density function is obtained.
- the Gaussian kernel density estimation can handle data distribution of any shape and has strong flexibility and adaptability.
- the choice of bandwidth has an important impact on the accuracy of the estimation result, and the result of the kernel density estimation can also be made more accurate by changing the bandwidth calculation method.
- the step S400: using a kernel density estimation method to calculate the second neural electrical pulse signal sample to obtain a second probability density function specifically includes:
- the calculation formula of the kernel density estimation method is:
- the kernel function K is a linear kernel xi is the sample point, n is the number of sample points, h is the bandwidth, and the bandwidth is
- the second neural electrical pulse signal sample is calculated using a linear kernel density estimation method to obtain a second probability density function.
- the linear kernel density estimation method has a relatively simple operation process and a fast operation speed.
- the step S400: using a kernel density estimation method to calculate the second neural electrical pulse signal sample to obtain a second probability density function specifically includes:
- the calculation formula of the kernel density estimation method is:
- the kernel function K is the cosine kernel
- ⁇ is the circumference of a circle
- xi is the sample point
- n is the number of sample points
- h is the bandwidth
- the bandwidth is the bandwidth
- the second neural electrical pulse signal sample is calculated using a triangular kernel density estimation method to obtain a second probability density function.
- the kernel density estimation method may also use a polynomial kernel, a Laplace kernel, a sigmoid kernel, a triangular kernel, or an exponential kernel, which may be selected according to needs and will not be described in detail here.
- Step S500 Calculate the information divergence between the first probability density function and the second probability density function, determine whether the first neural electrical pulse signal sample belongs to a neural electrical pulse according to the information divergence, and output the determination result.
- the first neural electrical pulse signal sample belongs to a neural electrical pulse, which has good specificity, does not lose a large amount of waveform information, and avoids the complex process of template matching.
- the step S500 calculating the information divergence between the first probability density function and the second probability density function, judging whether the first neural electrical pulse signal sample belongs to a neural electrical pulse according to the information divergence, and outputting the judgment result, specifically includes:
- KDE(i) represents the second probability density function
- KL(SN,KDE) represents the information divergence
- the judgment result is that the first neural electrical pulse signal sample belongs to a neural electrical pulse.
- the probability density of the first probability density parameter (SN-pdf) and the second probability density function (KDE-pdf) between the first neural electrical pulse signal sample value and the second neural electrical pulse signal sample value that is, the probability density between the maximum value max(Y) of Y and the minimum value min(Y) of Y, is calculated respectively, and the information divergence between the first probability density parameter and the second probability density function is calculated.
- the calculation formula of the information divergence is:
- SN(i) represents the first probability density parameter
- KDE(i) represents the second probability density parameter
- KL(SN, KDE) represents the information divergence, that is, the inconsistency or similarity of the two distributions. The more similar the two distributions are, the closer KL is to 0, and vice versa.
- calculating the KL divergence of two distributions is essentially a method of calculating entropy, which can also be changed to other entropy measure.
- the first neural electrical pulse signal sample belongs to a neural electrical pulse, compare the calculated information divergence KL with the preset threshold Th, if the information divergence KL is greater than the preset threshold Th, then the first neural electrical pulse signal sample is denied, if the information divergence is less than or equal to the preset threshold, then the first neural electrical pulse signal sample is accepted, and the judgment result is that the first neural electrical pulse signal sample belongs to a neural electrical pulse.
- the feature extraction process requires researchers to select appropriate features based on their own engineering practice experience, such as statistical features (mean, variance and quantile, etc.), spectral features (energy and spectral density, etc.), entropy features (sample entropy, differential entropy and fuzzy entropy, etc.) and other nonlinear features (fractal dimension and Lyapunov index, etc.).
- statistical features mean, variance and quantile, etc.
- spectral features energy and spectral density, etc.
- entropy features example entropy, differential entropy and fuzzy entropy, etc.
- other nonlinear features fractal dimension and Lyapunov index, etc.
- Template matching requires enough templates to be prepared in advance to minimize the missed detection rate. If it is assumed that the neural electrical pulse signals emitted by each neuron are different, then the brain has a total of about 80 billion neurons, which is a number that cannot be prepared in advance.
- the method of extracting templates online requires continuous learning so that the template is updated to a more stable state, and this process will cause the time cost to rise rapidly.
- this application does not require a large number of samples to be prepared in advance.
- There are two common solutions for extracting features using neural networks one is a supervised autoencoder and the other is an unsupervised automatic codec.
- the former requires a large number of manually labeled samples so that the neural network can learn a coding model that automatically extracts neural electrical pulse features.
- the latter in the process of learning a large amount of data, learns a model that maximizes the difference between samples of different classes and minimizes the feature differences of samples of the same class.
- This model can be used for sorting neural electrical pulses, but is rarely used as a detection tool for neural electrical pulses.
- Figure 13 is a schematic diagram of the principle of a preferred embodiment of the neural electrical pulse detection system of the present application
- Figure 14 is a schematic diagram of the operating environment of a preferred embodiment of the terminal of the present application.
- the present application further proposes a neural electrical pulse detection system, the neural electrical pulse detection system comprising:
- the signal preprocessing module 51 is used to obtain the original neural electrophysiological signal, filter and threshold the original neural electrophysiological signal, and obtain a first neural electrical pulse signal sample;
- a signal moment estimation module 52 configured to make a skewed distribution assumption for the first neural electrical pulse signal sample, obtain a first probability density parameter, a position parameter and a scale parameter of the skewed distribution of the first neural electrical pulse signal sample according to the moment estimation of the skewed distribution, and obtain a first probability density function according to the first probability density parameter, the position parameter and the scale parameter;
- a parameter determination module 53 configured to determine whether the first probability density parameter is less than zero, and if so, to use the first neural electrical pulse signal sample as a second neural electrical pulse signal sample;
- the signal density estimation module 54 is used to estimate the second neural electrical pulse signal sample using a kernel density estimation method. Calculate and obtain the second probability density function;
- the signal judgment module 55 is used to calculate the information divergence between the first probability density function and the second probability density function, judge whether the first neural electrical pulse signal sample belongs to a neural electrical pulse according to the information divergence, and output the judgment result.
- the present application also proposes a terminal, which includes: a memory 20, a processor 10, and a display 30.
- FIG14 only shows some components of the terminal, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.
- the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory of the terminal. In other embodiments, the memory 20 may also be an external storage device of the terminal, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the terminal. Further, the memory 20 may also include both an internal storage unit of the terminal and an external storage device. The memory 20 is used to store application software and various types of data installed in the terminal, such as the program code of the installation terminal, etc. The memory 20 may also be used to temporarily store data that has been output or is to be output.
- a plug-in hard disk such as a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the terminal.
- the memory 20 may also include both an internal storage unit of the terminal and an external storage device.
- the memory 20
- a neural electrical pulse detection program 40 is stored in the memory 20, and the neural electrical pulse detection program 40 can be executed by the processor 10, thereby realizing the neural electrical pulse detection method in the present application.
- the processor 10 can be a central processing unit (CPU), a microprocessor or other data processing chip, which is used to run the program code or process data stored in the memory 20, such as executing the neural electrical pulse detection method.
- CPU central processing unit
- microprocessor or other data processing chip, which is used to run the program code or process data stored in the memory 20, such as executing the neural electrical pulse detection method.
- the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch device, etc.
- the display 30 is used to display information on the terminal and to display a visual user interface.
- the components 10-30 of the terminal communicate with each other via a system bus.
- the present application also proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a neural electrical pulse detection program, and when the neural electrical pulse detection program is executed by a processor, the steps of the neural electrical pulse detection method as described above are implemented.
- the present application obtains original neural electrophysiological signals from animal or human brains, performs filtering and threshold detection on the original neural electrophysiological signals, thereby reducing the noise of the original neural electrophysiological signals and obtaining a first neural electrical pulse signal sample, thereby improving the anti-noise ability;
- the present application makes a skewed distribution assumption on the first neural electrical pulse signal sample, and obtains the first probability density parameter, position parameter and scale parameter of the skewed distribution of the first neural electrical pulse signal sample according to the moment estimation of the skewed distribution, so as to facilitate the use of the density estimation algorithm and the judgment of whether the first neural electrical pulse signal sample belongs to a neural electrical pulse in subsequent steps;
- the present application compares whether the first probability density parameter satisfies ⁇ 0, and if so, obtains the second neural electrical pulse signal sample, and obtains a qualified second neural electrical pulse signal sample after an initial judgment, thereby avoiding the first probability density parameter that does not meet the conditions from entering the next step; in addition, the present
- this invention determines whether the first neural electrical pulse signal sample belongs to a neural electrical pulse by calculating the information divergence between the first probability density parameter and the second probability density function. It has good specificity, does not lose a large amount of waveform information, and avoids the complex process of template matching.
- the terms “installed”, “connected”, “connected”, etc. should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal communication of two components.
- installed can be a fixed connection, a detachable connection, or an integral connection
- it can be a mechanical connection or an electrical connection
- it can be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal communication of two components.
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Veterinary Medicine (AREA)
- Public Health (AREA)
- Pathology (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Signal Processing (AREA)
- Psychiatry (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physiology (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Psychology (AREA)
- Neurosurgery (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Neurology (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
- Measuring Pulse, Heart Rate, Blood Pressure Or Blood Flow (AREA)
Abstract
一种神经电脉冲检测方法、系统及终端,方法包括:获取原始神经电生理信号,对原始神经电生理信号通滤波和阈值检测,得到第一神经电脉冲信号样本(S100);对第一神经脉冲信号作偏态分布假设,根据偏态分布的矩估计得到第一神经电脉冲信号样本的偏态分布的第一概率密度参数、位置参数和尺度参数,并根据第一概率密度参数、位置参数和尺度参数得到第一概率密度函数(S200);判断第一概率密度参数是否小于零,若是,将第一神经电脉冲信号样本作为第二神经电脉冲信号样本(S300);使用核密度估计法对第二神经电脉冲信号样本计算,得到第二概率密度函数(S400);计算第一概率密度函数和第二概率密度函数之间的信息散度,根据信息散度判断第一神经电脉冲信号样本是否属于神经电脉冲,并输出判断结果(S500),具有快速高效、抗噪声能力高、特异性好,无需模板匹配的优点。
Description
本申请涉及人工智能技术领域,特别是涉及一种神经电脉冲检测方法、系统、终端及计算机可读存储介质。
神经电活动的研究是人类了解和认识大脑的重要途径、是研发前沿脑机接口的重要抓手,对人类生命健康和国家经济发展均具有重要的价值与意义。神经电脉冲检测是神经科学和生物医学工程领域的核心课题,它涉及从复杂的脑电信号中准确、高效地识别出神经元的放电活动。技术的进步尤其是高通量电极阵列的发展为这一任务提出了新的挑战,即如何克服从成百上千的电极通道中快速准确检测到稀疏神经电脉冲的难题。
在现有技术中,有些研究是基于研究人员获得的神经电脉冲的先验知识,人为设置阈值进行检测,这种方法具有低抗噪声能力或较差的特异性;有些研究是基于特征的检测,即在阈值检测的基础上,提取包括但不限于时域、频域、熵、非线性动力学等测度作为检测判断电脉冲信号的特征,这种方法的缺点是,特征提取属于降维的方法,过程中不可避免地会损失大量波形信息;有些研究是基于模板匹配,模板匹配是通过计算待定波形与模板波形之间的相似度实现神经电脉冲检测,这个过程中模板的来源和个数是影响检测效率和准确率的重要因素,相对麻烦且不易控制;因此,如何找到快速高效的神经电脉冲检测方法是目前需要解决的问题。
发明内容
有鉴于此,本申请提供了一种神经电脉冲检测方法、系统、终端及计算机可读存储介质,以解决现有技术中的神经电脉冲检测方法,抗噪声能力低,特异性差,会损失大量波形信息,模板匹配过程复杂的问题。
本申请提出一种神经电脉冲检测方法,所述神经电脉冲检测方法包括:
获取原始神经电生理信号,对所述原始神经电生理信号进行通滤波和阈值检测,得到第一神经电脉冲信号样本;
对所述第一神经电脉冲信号样本作偏态分布假设,根据偏态分布的矩估计得到所述第一神经电脉冲信号样本的偏态分布的第一概率密度参数、位置参数和尺度参数,并根据所述第一概率密度参数、所述位置参数和所述尺度参数得到第一概率密度函数;
判断所述第一概率密度参数是否小于零,若是,将所述第一神经电脉冲信号样本作为第二神经电脉冲信号样本;
使用核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数;
计算所述第一概率密度函数和所述第二概率密度函数之间的信息散度,根据所述信息散度判断所述第一神经电脉冲信号样本是否属于神经电脉冲,并输出判断结果。
可选地,所述获取原始神经电生理信号,对所述原始神经电生理信号进行通滤波和阈值检测,得到第一神经电脉冲信号样本,具体包括:
获取原始神经电生理信号,使用六阶巴特沃斯滤波器对所述原始神经电生理信号进行通滤波,得到第一滤波电脉冲信号;
使用阈值检测参数对所述第一滤波电脉冲信号进行阈值检测,得到所述第一神经电脉冲信号样本。
可选地,所述对所述第一神经电脉冲信号样本作偏态分布假设,根据偏态分布的矩估计得到所述第一神经电脉冲信号样本的偏态分布的第一概率密度参数、位置参数和尺度参数,并根据所述第一概率密度参数、所述位置参数和所述尺度参数得到第一概率密度函数,具体包括:
对所述第一神经电脉冲信号样本作偏态分布假设:
Y~SN(ξ,ω2,α);
Y~SN(ξ,ω2,α);
其中,Y表示所述第一神经电脉冲信号样本,SN表示偏态分布;
Y=ξ+ωX;
Y=ξ+ωX;
其中,ξ∈R为位置参数,ω∈R+为尺度参数,α为第一概率密度参数,α<0为倾斜变量,X为服从标准偏态分布样本;
根据偏态分布的矩估计得到所述第一神经电脉冲信号样本的偏态分布的第一概率密度参数、位置参数和尺度参数,并根据所述第一概率密度参数、位置参数和尺度参数得到第一概率密度函数;
所述第一概率密度函数的数学公式为:
其中,表示标准高斯密度函数,Φ(α·x)为累积概率分布函数,x为随机变量。
可选地,所述使用核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数,具体包括:
使用高斯核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数;
核密度估计法的计算公式为:
其中,核函数K为高斯核xi为样本点,n为样本点个数,h为带宽,带宽为
可选地,所述使用核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数,具体包括:
使用线性核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数;
核密度估计法的计算公式为:
其中,核函数K为线性核xi为样本点,n为样本点个数,h为带宽,带宽为
可选地,所述使用核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数,具体包括:
使用余弦核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数;
核密度估计法的计算公式为:
其中,核函数K为余弦核其中,π为圆周率,xi为样本点,n为样本点个数,h为带宽,带宽为
可选地,所述计算所述第一概率密度函数和所述第二概率密度函数之间的信息散度,根据所述信息散度判断所述第一神经电脉冲信号样本是否属于神经电脉冲,并输出判断结果,具体包括:
分别计算所述第一概率密度函数和所述第二概率密度函数在所述第一神经电脉冲信号样本值和第二神经电脉冲信号样本值之间的概率密度;
根据所述第一神经电脉冲信号样本值和所述第二神经电脉冲信号样本值之间的概率密度计算所述第一概率密度函数和所述第二概率密度函数之间的信息散度;
所述信息散度的计算公式为:
其中,SN(i)表示第一概率密度函数,KDE(i)表示第二概率密度函数,KL(SN,KDE)表示信息散度;
判断所述第一神经电脉冲信号样本是否属于神经电脉冲,并输出判断结果;
若信息散度小于或者等于预设阈值,则所述判断结果为所述第一神经电脉冲信号样本属于神经电脉冲。
本申请还提出一种神经电脉冲检测系统,所述神经电脉冲检测系统包括:
信号预处理模块,用于获取原始神经电生理信号,对所述原始神经电生理信号进行通滤波和阈值检测,得到第一神经电脉冲信号样本;
信号矩估计模块,用于对所述第一神经电脉冲信号样本作偏态分布假设,根据偏态分布的矩估计得到所述第一神经电脉冲信号样本的偏态分布的第一概率密度参数、位置参数和尺度参数,并根据所述第一概率密度参数、所述位置参数和所述尺度参数得到第一概率密度函数;
参数判断模块,用于判断所述第一概率密度参数是否小于零,若是,将所述第一神经电脉冲信号样本作为第二神经电脉冲信号样本;
信号密度估计模块,用于使用核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数;
信号判断模块,用于计算所述第一概率密度函数和所述第二概率密度函数之间的信息
散度,根据所述信息散度判断所述第一神经电脉冲信号样本是否属于神经电脉冲,并输出判断结果。
本申请还提出一种终端,所述终端包括:存储器、处理器及存储在所述存储器上并可在所述处理器上运行的神经电脉冲检测程序,所述神经电脉冲检测程序被所述处理器执行时实现如所述的神经电脉冲检测方法的步骤。
本申请还提出一种计算机可读存储介质,所述计算机可读存储介质存储有神经电脉冲检测程序,所述神经电脉冲检测程序被处理器执行时实现如所述的神经电脉冲检测方法的步骤。
本申请的有益效果是:区别于现有技术,本申请通过获取原始神经电生理信号,对原始神经电生理信号进行通滤波和阈值检测,使得对原始神经电生理信号进行降噪,得到第一神经电脉冲信号样本,提高了抗噪声能力;其次,本申请通过对第一神经电脉冲信号样本作偏态分布假设,根据偏态分布的矩估计得到第一神经电脉冲信号样本的偏态分布的第一概率密度参数、位置参数和尺度参数,并根据所述第一概率密度参数、位置参数和尺度参数得到第一概率密度函数,便于后续步骤中使用密度估计算法和判断第一神经电脉冲信号样本是否属于神经电脉冲;再次,本申请通过判断所述第一概率密度参数是否小于零,若是,将所述第一神经电脉冲信号样本作为第二神经电脉冲信号样本,避免不满足条件的第一概率密度参数进入下一个步骤;此外,本申请对第二神经电脉冲信号样本使用核密度估计法进行计算,得到第二概率密度函数,能够快速高效的进行神经电脉冲检测;另外,本申请通过计算所述第一概率密度函数和所述第二概率密度函数之间的信息散度,根据所述信息散度判断所述第一神经电脉冲信号样本是否属于神经电脉冲,并输出判断结果,特异性好,不会损失大量波形信息,且避免了模板匹配的复杂过程。
应当理解的是,以上的一般描述和后文的细节描述仅是示例性和解释性的,而非限制本申请。
为了更清楚地说明本申请实施例中的技术方案,下面将对实施例描述中所需要使用的附图作简单地介绍,显而易见地,下面描述中的附图仅仅是本申请的一些实施例,对于本领域普通技术人员来讲,在不付出创造性劳动的前提下,还可以根据这些附图获得其他的附图。
图1是本申请神经电脉冲检测方法的较佳实施例的流程图;
图2是本申请神经电脉冲检测方法中的原始神经电生理信号图;
图3是本申请神经电脉冲检测方法中的第一神经电脉冲信号样本图;
图4是本申请神经电脉冲检测方法中的第一类大量神经电脉冲信号样本;
图5是本申请神经电脉冲检测方法中的第二类第一神经电脉冲信号样本图;
图6是本申请神经电脉冲检测方法中的第一类噪声样本的统计直方图;
图7是本申请神经电脉冲检测方法中的第二类噪声样本的统计直方图;
图8是本申请神经电脉冲检测方法中的第三类噪声样本的统计直方图;
图9是本申请神经电脉冲检测方法中的第四类噪声样本的统计直方图;
图10是本申请神经电脉冲检测方法中的单个神经电脉冲信号的直方图;
图11是本申请神经电脉冲检测方法中的样本理论SN概率密度分布图;
图12是本申请神经电脉冲检测方法中的高斯核密度估计后的概率密度分布图;
图13是本申请神经电脉冲检测系统的较佳实施例的原理示意图;
图14为本申请终端的较佳实施例的运行环境示意图。
为使本领域的技术人员更好地理解本申请的技术方案,下面结合附图和具体实施方式对本申请所提供的神经电脉冲检测方法、系统、终端及计算机可读存储介质做进一步详细描述。可以理解的是,所描述的实施例仅仅是本申请一部分实施例,而不是全部实施例。基于本申请中的实施例,本领域普通技术人员在没有做出创造性的劳动前提下所获得的所有其他实施例,都属于本申请保护的范围。
本申请中的术语“第一”、“第二”等是用于区别不同对象,而不是用于描述特定顺序。此外,术语“包括”和“具有”以及它们任何变形,意图在于覆盖不排他的包含。例如包含了一系列步骤或单元的过程、方法、系统、产品或设备没有限定于已列出的步骤或单元,而是可选地还包括没有列出的步骤或单元,或可选地还包括对于这些过程、方法、产品或设备固有的其他步骤或单元。
本申请提供一种神经电脉冲检测方法、系统、终端及计算机可读存储介质,以解决现有技术中的神经电脉冲检测方法,抗噪声能力低,特异性差,会损失大量波形信息,模板匹配过程复杂的问题。
请参阅图1至图12,图1是本申请神经电脉冲检测方法的较佳实施例的流程图;图2是本申请神经电脉冲检测方法中的原始神经电生理信号图;图3是本申请神经电脉冲检测方法中的第一神经电脉冲信号样本图;图4是本申请神经电脉冲检测方法中的第一类大量神经电脉冲信号样本;图5是本申请神经电脉冲检测方法中的第二类第一神经电脉冲信号样本图;图6是本申请神经电脉冲检测方法中的第一类噪声样本的统计直方图;图7是本申请神经电脉冲检测方法中的第二类噪声样本的统计直方图;图8是本申请神经电脉冲检测方法中的第三类噪声样本的统计直方图;图9是本申请神经电脉冲检测方法中的第四类噪声样本的统计直方图;图10是本申请神经电脉冲检测方法中的单个神经电脉冲信号的直方图;图11是本申请神经电脉冲检测方法中的样本理论SN概率密度分布图;图12是本申请神经电脉冲检测方法中的高斯核密度估计后的概率密度分布图。
本申请提出一种神经电脉冲检测方法,其中,如图1所示,神经电脉冲检测方法包括步骤:
步骤S100:获取原始神经电生理信号,对所述原始神经电生理信号进行通滤波和阈值检测,得到第一神经电脉冲信号样本。
具体的,获取原始神经电生理信号,对原始神经电生理信号进行通滤波和阈值检测,对原始神经电生理信号进行降噪,得到第一神经电脉冲信号样本,提高了抗噪声能力。
所述步骤S100:获取原始神经电生理信号,对所述原始神经电生理信号进行通滤波和阈值检测,得到第一神经电脉冲信号样本,具体包括:
获取原始神经电生理信号,使用六阶巴特沃斯滤波器对所述原始神经电生理信号进行通滤波,得到第一滤波电脉冲信号。
其中,所述六阶巴特沃斯滤波器的第一截止频率和第二截止频率分别为250赫兹和7000赫兹。
使用阈值检测参数对所述第一滤波电脉冲信号进行阈值检测,得到所述第一神经电脉冲信号样本。
其中,所述阈值检测参数设置为信号的负向峰值大于60微伏特,且峰峰间距设置为大于2毫秒,并以峰值时刻为基点,向前取1毫秒,向后取1.5毫秒视为所述第一神经电脉冲信号样本。
具体的,如图2所示,采集到的信号为原始神经电生理信号,获取原始神经电生理信号,在这些信号序列中通常包含了背景噪声和研究人员关注的神经电脉冲信号,对原始神经电生理信号使用前后向六阶巴特沃斯滤波器进行通滤波,得到第一滤波电脉冲信号,其中,六阶巴特沃斯滤波器的第一截止频率和第二截止频率分别为250赫兹和7000赫兹,第一截止频率为上限截止频率,第二截止频率为下限截止频率。并使用阈值检测参数对第一滤波电脉冲信号进行阈值检测,得到第一神经电脉冲信号样本。
其中,阈值检测参数设置为信号的负向峰值大于60微伏特,且峰峰间距设置为大于2毫秒,并以峰值时刻为基点,向前取1毫秒,向后取1.5毫秒视为第一神经电脉冲信号样本,阈值检测参数设置为信号的负向峰值大于60微伏特,且峰峰间距大于2毫秒,如图3所示,并以峰值时刻为基点,向前取1毫秒,向后取1.5毫秒视为潜在的第一神经电脉冲信号样本。
步骤S200:对所述第一神经电脉冲信号样本作偏态分布假设,根据偏态分布的矩估计得到所述第一神经电脉冲信号样本的偏态分布的第一概率密度参数、位置参数和尺度参数,并根据所述第一概率密度参数、所述位置参数和所述尺度参数得到第一概率密度函数。
具体的,由于第一神经电脉冲信号样本的分布曲线属于左右不对称的数据次数分布,更符合偏态分布,因此,对第一神经电脉冲信号样本作偏态分布假设,根据偏态分布的矩估计得到第一神经电脉冲信号样本的偏态分布的第一概率密度参数、位置参数和尺度参数。
需要说明的是,本申请对神经电脉冲的SN分布假设来自于对大量样本的统计推断。如图4和图5所示,分别为两种类别神经电脉冲信号的统计直方图,可以看出不同类别的神经电脑部信号虽然在大小不同,但是都属于SN分布,且其分布参数为α为小于0。
如图6、图7、图8和图9所示,不同类别的噪声样本的统计直方图,可以看出,不同类别的噪声样本的统计直方图表现不一,且非α<0的SN分布。
其中,所述步骤S200:对所述第一神经电脉冲信号样本作偏态分布假设,根据偏态分布的矩估计得到所述第一神经电脉冲信号样本的偏态分布的第一概率密度参数、位置参数和尺度参数,并根据所述第一概率密度参数、所述位置参数和所述尺度参数得到第一概率密度函数,具体包括:
对所述第一神经电脉冲信号样本作偏态分布假设:
Y~SN(ξ,ω2,α);
Y~SN(ξ,ω2,α);
其中,Y表示所述第一神经电脉冲信号样本,SN表示偏态分布;
Y=ξ+ωX;
Y=ξ+ωX;
其中,ξ∈R为位置参数,ω∈R+为尺度参数,α为第一概率密度参数,α<0为倾斜变量,X为服从标准偏态分布样本;
根据偏态分布的矩估计得到所述第一神经电脉冲信号样本的偏态分布的第一概率密度参数、位置参数和尺度参数,并根据所述第一概率密度参数、位置参数和尺度参数得到第一概率密度函数;
所述第一概率密度函数的数学公式为:
其中,表示标准高斯密度函数,Φ(α·x)为累积概率分布函数,x为随机变量。
具体的,对第一神经电脉冲信号样本作偏态分布(Skew-Normal distribution,SN分布)假设,可得到公式:Y~SN(ξ,ω2,α),
其中,Y表示第一神经电脉冲信号样本,SN表示偏态分布。
进一步,则有公式:Y=ξ+ωX;
其中,ξ∈R为位置参数,ω∈R+为尺度参数,α为第一概率密度参数,α<0为倾斜变量,X为服从标准偏态分布样本;
根据偏态分布的矩估计得到第一神经电脉冲信号样本的偏态分布的第一概率密度参数;
其中,概率密度函数数学公式为:其中,表示标准高斯密度函数,Φ(α·x)为累积概率分布函数,x为随机变量。
在上述前提下,根据偏态概率密分布的矩估计得到第一神经电脉冲信号样本Y的SN分布概率密度参数,从而得到理论上第一神经电脉冲信号样本Y的概率密度函数,即SN-pdf。
如图10、图11和图12所示,单个神经电脉冲样本的估计图,可以看出,对于单个神经电脉冲样本的估计和利用高斯核密度估计也符合预期。如图10所示,为单个神经电脉冲信号的直方图,如图11所示,为该样本理论SN概率密度分布,如图12所示,为采用高斯核密度估计后的概率密度分布。这些实验结果说明采用概率密度估计的方法可以作为一种快速高效的神经电脉冲检测方法。
步骤S300:判断所述第一概率密度参数是否小于零,若是,将所述第一神经电脉冲信号样本作为第二神经电脉冲信号样本。
具体的,判断得到的理论SN分布第一概率密度参数是否小于零,即是否满足α<0的条件,若是,得到第二神经电脉冲信号样本,若否,则拒绝接受第一神经电脉冲信号样本。
步骤S400:使用核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数。
具体的,对第二神经电脉冲信号样本使用核密度估计法进行计算,得到第二概率密度函数,能够快速高效的进行神经电脉冲检测。
其中,所述步骤S400:使用核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数,具体包括:
使用高斯核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数;
核密度估计法的计算公式为:
其中,核函数K为高斯核xi为样本点,n为样本点个数,h为带宽,带宽为
具体的,利用核密度估计的思想,估计第二神经电脉冲信号样本的真实概率密度分布(KDE-pdf),使用高斯核密度估计法进行计算,高斯核密度估计法进行计算公式为:
其中,核函数为高斯核xi为样本点,n为样本点个数,h为带宽,带宽为得到第二概率密度函数,高斯核密度估计可以处理任意形状的数据分布,具有较强的灵活性和适应性。
可选地,带宽的选择对估计结果的精度有重要影响,也可以通过改变带宽的计算方法使得核密度估计的结果更精确。
或者,所述步骤S400:使用核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数,具体包括:
使用线性核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数;
核密度估计法的计算公式为:
其中,核函数K为线性核xi为样本点,n为样本点个数,h为带宽,带宽为
具体的,对第二神经电脉冲信号样本使用线性核密度估计法进行计算,得到第二概率密度函数,线性核密度估计法运算过程相对简单,运算速度快。
或者,所述步骤S400:使用核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数,具体包括:
使用余弦核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数;
核密度估计法的计算公式为:
其中,核函数K为余弦核其中,π为圆周率,xi为样本点,n为样本点个数,h为带宽,带宽为
具体的,对第二神经电脉冲信号样本使用三角核密度估计法进行计算,得到第二概率密度函数。
可选地,核密度估计法还可以使用多项式核、拉普拉斯核、sigmoid核、三角核、指数核,具体可根据需要进行选择,此处不再一一赘述。
步骤S500:计算所述第一概率密度函数和所述第二概率密度函数之间的信息散度,根据所述信息散度判断所述第一神经电脉冲信号样本是否属于神经电脉冲,并输出判断结果。
具体的,通过计算第一概率密度参数和第二概率密度函数之间的信息散度,判断第一神经电脉冲信号样本是否属于神经电脉冲,特异性好,不会损失大量波形信息,且避免了模板匹配的复杂过程。
其中,所述步骤S500:计算所述第一概率密度函数和所述第二概率密度函数之间的信息散度,根据所述信息散度判断所述第一神经电脉冲信号样本是否属于神经电脉冲,并输出判断结果,具体包括:
分别计算所述第一概率密度函数和所述第二概率密度函数在所述第一神经电脉冲信号样本值和第二神经电脉冲信号样本值之间的概率密度;
根据所述第一神经电脉冲信号样本值和所述第二神经电脉冲信号样本值之间的概率密度计算所述第一概率密度函数和所述第二概率密度函数之间的信息散度;
所述信息散度的计算公式为:
其中,SN(i)表示第一概率密度函数,KDE(i)表示第二概率密度函数,KL(SN,KDE)表示信息散度;
判断所述第一神经电脉冲信号样本是否属于神经电脉冲,并输出判断结果;
若信息散度小于或者等于预设阈值,则所述判断结果为所述第一神经电脉冲信号样本属于神经电脉冲。
具体的,分别计算第一概率密度参数(SN-pdf)和第二概率密度函数(KDE-pdf)在第一神经电脉冲信号样本值和第二神经电脉冲信号样本值之间的概率密度,即在Y的最大值max(Y)和Y的最小值min(Y)之间的概率密度,并计算第一概率密度参数和第二概率密度函数之间的信息散度,信息散度的计算公式为:
其中,SN(i)表示第一概率密度参数,KDE(i)表示第二概率密度参数,KL(SN,KDE)表示信息散度,即表示两种分布的不一致性,或者相似性。两种分布越相似,则KL越接近于0,反之则越大;
可选地,计算两种分布的KL散度本质上为一种熵的计算方法,也可以更改为其它熵
测度。
最后,判断第一神经电脉冲信号样本是否属于神经电脉冲,对计算得到的信息散度KL与预设的阈值Th进行比较,若信息散度KL大于预设阈值Th,则否定第一神经电脉冲信号样本,若信息散度小于或者等于预设阈值,则接受第一神经电脉冲信号样本,所述判断结果为所述第一神经电脉冲信号样本属于神经电脉冲。
本申请相较于特征提取方案,可以减少对研究者工程经验的依赖。特征提取过程需要研究人员根据自己的工程实践经验选择合适的特征,如统计特征(均值、方差和分位数等)、频谱特征(能量和谱密度等)、熵特征(样本熵、微分熵和模糊熵等)和其它非线性特征(分形维数和李雅普诺夫指数等)。如何选择或组合上述特征进行神经电脉冲信号的检测通常是工程开发者决定的,这样的决策过程往往取决于开发者所熟悉特的征提取技术和所了解的神经电脉冲的特点。这种方案最终得到的神经电脉冲检测器通常不具备较强的鲁棒性。
本申请相较于模板匹配的方案,不需要提前准备合适的模板,也不需要在检测过程中提取模板(在线提取模板)。模板匹配需要提前准备足够的模板,才能最小化漏检率。如果假设每个神经元所发放的神经电脉冲信号都不尽相同,那么大脑共计约800亿个神经元,这是无法提前准备的一个数字。而在线提取模板的方法,则需要不断的学习,从而使得模板更新到一个较为稳定的状态,该过程会使得时间成本快速上升。
本申请相较于使用神经网络模型自动提取特征的方案,不需要提前准备大量的样本。神经网络提取特征的常见方案有两种,一种为有监督的自动编码器,另一种是无监督的自动编解码器。前一种需要大量的人工标注样本,从而使得神经网络学习到自动提取神经电脉冲特征的编码模型。后一种则在对大量数据的学习过程中,学到区别最大化不同类样本和最小化同类样本特征差异的模型,该模型可以用于神经电脉冲的排序工作,较少利用其作为神经电脉冲的检测工具。
请参阅图13至图14,图13是本申请神经电脉冲检测系统的较佳实施例的原理示意图;图14为本申请终端的较佳实施例的运行环境示意图。
在一些实施例中,如图13所示,基于上述神经电脉冲检测方法,本申请还提出一种神经电脉冲检测系统,所述神经电脉冲检测系统包括:
信号预处理模块51,用于获取原始神经电生理信号,对所述原始神经电生理信号进行通滤波和阈值检测,得到第一神经电脉冲信号样本;
信号矩估计模块52,用于对所述第一神经电脉冲信号样本作偏态分布假设,根据偏态分布的矩估计得到所述第一神经电脉冲信号样本的偏态分布的第一概率密度参数、位置参数和尺度参数,并根据所述第一概率密度参数、所述位置参数和所述尺度参数得到第一概率密度函数;
参数判断模块53,用于判断所述第一概率密度参数是否小于零,若是,将所述第一神经电脉冲信号样本作为第二神经电脉冲信号样本;
信号密度估计模块54,用于使用核密度估计法对所述第二神经电脉冲信号样本进行
计算,得到第二概率密度函数;
信号判断模块55,用于计算所述第一概率密度函数和所述第二概率密度函数之间的信息散度,根据所述信息散度判断所述第一神经电脉冲信号样本是否属于神经电脉冲,并输出判断结果。
在一些实施例中,如图14所示,基于上述神经电脉冲检测方法和系统,本申请还相应提出一种终端,所述终端包括:存储器20、处理器10、显示器30,图14仅示出了终端的部分组件,但是应理解的是,并不要求实施所有示出的组件,可以替代的实施更多或者更少的组件。
所述存储器20在一些实施例中可以是所述终端的内部存储单元,例如终端的硬盘或内存。所述存储器20在另一些实施例中也可以是所述终端的外部存储设备,例如所述终端上配备的插接式硬盘,智能存储卡(Smart Media Card,SMC),安全数字(Secure Digital,SD)卡,闪存卡(Flash Card)等。进一步地,所述存储器20还可以既包括所述终端的内部存储单元也包括外部存储设备。所述存储器20用于存储安装于所述终端的应用软件及各类数据,例如所述安装终端的程序代码等。所述存储器20还可以用于暂时地存储已经输出或者将要输出的数据。
在一实施例中,存储器20上存储有神经电脉冲检测程序40,该神经电脉冲检测程序40可被处理器10所执行,从而实现本申请中神经电脉冲检测方法。
所述处理器10在一些实施例中可以是一中央处理器(Central Processing Unit,CPU),微处理器或其他数据处理芯片,用于运行所述存储器20中存储的程序代码或处理数据,例如执行所述神经电脉冲检测方法等。
所述显示器30在一些实施例中可以是LED显示器、液晶显示器、触控式液晶显示器以及OLED(Organic Light-Emitting Diode,有机发光二极管)触摸器等。所述显示器30用于显示在所述终端的信息以及用于显示可视化的用户界面。所述终端的部件10-30通过系统总线相互通信。
本申请还提出一种计算机可读存储介质,所述计算机可读存储介质存储有神经电脉冲检测程序,所述神经电脉冲检测程序被处理器执行时实现如上所述的神经电脉冲检测方法的步骤。
综上所述,本申请通过从动物或人脑中获取原始神经电生理信号,对原始神经电生理信号进行通滤波和阈值检测,使得对原始神经电生理信号进行降噪,得到第一神经电脉冲信号样本,提高了抗噪声能力;其次,本申请通过对第一神经电脉冲信号样本作偏态分布假设,根据偏态分布的矩估计得到第一神经电脉冲信号样本的偏态分布的第一概率密度参数、位置参数和尺度参数,便于后续步骤中使用密度估计算法和判断第一神经电脉冲信号样本是否属于神经电脉冲;再次,本申请比较第一概率密度参数是否满足α<0,若是,得到第二神经电脉冲信号样本,经过初次判断,得到合格的第二神经电脉冲信号样本,避免不满足条件的第一概率密度参数进入下一个步骤;此外,本申请对第二神经电脉冲信号样
本使用核密度估计法进行计算,得到第二概率密度函数,能够快速高效的进行神经电脉冲检测;另外,本申请通过计算第一概率密度参数和第二概率密度函数之间的信息散度,判断第一神经电脉冲信号样本是否属于神经电脉冲,特异性好,不会损失大量波形信息,且避免了模板匹配的复杂过程。
需要说明的是,本申请实施例中介绍的多种可选的实施方式,彼此可以相互结合实现,也可以单独实现,对此本申请实施例不作限定。
在本申请的描述中,需要理解的是,术语“上”、“下”、“左”、“右”等指示方位或位置关系为基于附图所示的方位或位置关系,仅是为了便于描述本申请和简化描述,而不是指示或暗示所指的装置或元件必须具有特定的方位、以及特定的方位构造和操作。因此,不能理解为对本申请的限制。此外,“第一”、“第二”仅由于描述目的,且不能理解为指示或暗示相对重要性或者隐含指明所指示的技术特征的数量。因此,限定有“第一”、“第二”的特征可以明示或者隐含地包括一个或者多个该特征。本申请的描述中,除非另有说明,“多个”的含义是两个或两个以上。
在本申请的描述中,需要说明的是,除非另有明确的规定和限定,术语“安装”、“相连”、“连接”等应做广义理解,例如,可以是固定连接,也可以是可拆卸连接,或一体地连接;可以是机械连接,也可以是电连接;可以是直接相连,也可以通过中间媒介间接连接,可以是两个元件内部的连通。对于本领域的普通技术人员而言,可以具体情况理解上述术语在本申请中的具体含义。
上述实施例是参考附图来描述的,其他不同的形式和实施例也是可行而不偏离本申请的原理,因此本申请不应被建构成为在此所提出实施例的限制。更确切地说,这些实施例被提供以使得本申请会是完善又完整,且会将本申请范围传达给本领域技术人员。在附图中,组件尺寸及相对尺寸也许基于清晰起见而被夸大。在此所使用的术语只是基于描述特定实施例目的,并无意成为限制。术语“包含”及/或“包括”在使用于本说明书时,表示所述特征、整数、构件及/或组件的存在,但不排除一或更多其他特征整数、构件、组件及/或其族群的存在或增加。除非另有所示,陈述时,数值范围包含该范围的上下限及其间的任何子范围。
以上所述仅为本申请的部分实施例,并非因此限制本申请的保护范围,凡是利用本申请说明书及附图内容所作的等效装置或等效流程变换,或直接或间接运用在其他相关的技术领域,均同理包括在本申请的专利保护范围内。
Claims (10)
- 一种神经电脉冲检测方法,其特征在于,包括:获取原始神经电生理信号,对所述原始神经电生理信号进行通滤波和阈值检测,得到第一神经电脉冲信号样本;对所述第一神经电脉冲信号样本作偏态分布假设,根据偏态分布的矩估计得到所述第一神经电脉冲信号样本的偏态分布的第一概率密度参数、位置参数和尺度参数,并根据所述第一概率密度参数、所述位置参数和所述尺度参数得到第一概率密度函数;判断所述第一概率密度参数是否小于零,若是,将所述第一神经电脉冲信号样本作为第二神经电脉冲信号样本;使用核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数;计算所述第一概率密度函数和所述第二概率密度函数之间的信息散度,根据所述信息散度判断所述第一神经电脉冲信号样本是否属于神经电脉冲,并输出判断结果。
- 根据权利要求1所述的神经电脉冲检测方法,其特征在于,所述获取原始神经电生理信号,对所述原始神经电生理信号进行通滤波和阈值检测,得到第一神经电脉冲信号样本,具体包括:获取原始神经电生理信号,使用六阶巴特沃斯滤波器对所述原始神经电生理信号进行通滤波,得到第一滤波电脉冲信号;使用阈值检测参数对所述第一滤波电脉冲信号进行阈值检测,得到所述第一神经电脉冲信号样本。
- 根据权利要求1所述的神经电脉冲检测方法,其特征在于,所述对所述第一神经电脉冲信号样本作偏态分布假设,根据偏态分布的矩估计得到所述第一神经电脉冲信号样本的偏态分布的第一概率密度参数、位置参数和尺度参数,并根据所述第一概率密度参数、所述位置参数和所述尺度参数得到第一概率密度函数,具体包括:对所述第一神经电脉冲信号样本作偏态分布假设:
Y~SN(ξ,ω2,α);其中,Y表示所述第一神经电脉冲信号样本,SN表示偏态分布;
Y=ξ+ωX;其中,ξ∈R为位置参数,ω∈R+为尺度参数,α为第一概率密度参数,α<0为倾斜变量,X为服从标准偏态分布样本;根据偏态分布的矩估计得到所述第一神经电脉冲信号样本的偏态分布的第一概率密度参数、位置参数和尺度参数,并根据所述第一概率密度参数、所述位置参数和所述尺度参数得到第一概率密度函数;所述第一概率密度函数的数学公式为:其中,表示标准高斯密度函数,Φ(α·x)为累积概率分布函数,x为随机变量。 - 根据权利要求3所述的神经电脉冲检测方法,其特征在于,所述使用核密度估计 法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数,具体包括:使用高斯核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数;核密度估计法的计算公式为:其中,核函数K为高斯核xi为样本点,n为样本点个数,h为带宽,带宽为
- 根据权利要求3所述的神经电脉冲检测方法,其特征在于,所述使用核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数,具体包括:使用线性核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数;核密度估计法的计算公式为:其中,核函数K为线性核x<h,xi为样本点,n为样本点个数,h为带宽,带宽为
- 根据权利要求3所述的神经电脉冲检测方法,其特征在于,所述使用核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数,具体包括:使用余弦核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数;核密度估计法的计算公式为:其中,核函数K为余弦核x<h,其中,π为圆周率,xi为样本点,n为样本点个数,h为带宽,带宽为
- 根据权利要求1所述的神经电脉冲检测方法,其特征在于,所述计算所述第一概率密度函数和所述第二概率密度函数之间的信息散度,根据所述信息散度判断所述第一神经电脉冲信号样本是否属于神经电脉冲,并输出判断结果,具体包括:分别计算所述第一概率密度函数和所述第二概率密度函数在所述第一神经电脉冲信号样本值和第二神经电脉冲信号样本值之间的概率密度;根据所述第一神经电脉冲信号样本值和所述第二神经电脉冲信号样本值之间的概率密度计算所述第一概率密度函数和所述第二概率密度函数之间的信息散度;所述信息散度的计算公式为:其中,SN(i)表示第一概率密度函数,KDE(i)表示第二概率密度函数,KL(SN,KDE)表示信息散度;判断所述第一神经电脉冲信号样本是否属于神经电脉冲,并输出判断结果;若信息散度小于或者等于预设阈值,则所述判断结果为所述第一神经电脉冲信号样本属于神经电脉冲。
- 一种神经电脉冲检测系统,其特征在于,所述神经电脉冲检测系统包括:信号预处理模块,用于获取原始神经电生理信号,对所述原始神经电生理信号进行通滤波和阈值检测,得到第一神经电脉冲信号样本;信号矩估计模块,用于对所述第一神经电脉冲信号样本作偏态分布假设,根据偏态分布的矩估计得到所述第一神经电脉冲信号样本的偏态分布的第一概率密度参数、位置参数和尺度参数,并根据所述第一概率密度参数、所述位置参数和所述尺度参数得到第一概率密度函数;参数判断模块,用于判断所述第一概率密度参数是否小于零,若是,将所述第一神经电脉冲信号样本作为第二神经电脉冲信号样本;信号密度估计模块,用于使用核密度估计法对所述第二神经电脉冲信号样本进行计算,得到第二概率密度函数;信号判断模块,用于计算所述第一概率密度函数和所述第二概率密度函数之间的信息散度,根据所述信息散度判断所述第一神经电脉冲信号样本是否属于神经电脉冲,并输出判断结果。
- 一种终端,其特征在于,所述终端包括:存储器、处理器及存储在所述存储器上并可在所述处理器上运行的神经电脉冲检测程序,所述神经电脉冲检测程序被所述处理器执行时实现如权利要求1-7任一项所述的神经电脉冲检测方法的步骤。
- 一种计算机可读存储介质,其特征在于,所述计算机可读存储介质存储有神经电脉冲检测程序,所述神经电脉冲检测程序被处理器执行时实现如权利要求1-7任一项所述的神经电脉冲检测方法的步骤。
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311675784.3A CN117357134B (zh) | 2023-12-08 | 2023-12-08 | 一种神经电脉冲检测方法、系统及终端 |
| CN202311675784.3 | 2023-12-08 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2025118317A1 true WO2025118317A1 (zh) | 2025-06-12 |
Family
ID=89400692
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2023/138053 Pending WO2025118317A1 (zh) | 2023-12-08 | 2023-12-12 | 一种神经电脉冲检测方法、系统及终端 |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN117357134B (zh) |
| WO (1) | WO2025118317A1 (zh) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118013308B (zh) * | 2024-04-10 | 2024-06-04 | 中国科学院深圳先进技术研究院 | 基于自适应树的时空增强神经电脉冲信号聚类方法及系统 |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106529186A (zh) * | 2016-11-24 | 2017-03-22 | 浙江大学 | 一种脑区脉冲神经信号的预测方法 |
| WO2021249313A1 (zh) * | 2020-06-10 | 2021-12-16 | 上海海事大学 | 一种基于可变分组的偏态分布最优参数估计方法 |
| CN114469138A (zh) * | 2022-01-13 | 2022-05-13 | 博睿康医疗科技(上海)有限公司 | 基于时频域的脑电爆发抑制模式的检测方法、系统及介质 |
| CN114663468A (zh) * | 2020-12-04 | 2022-06-24 | 株式会社Ntt都科摩 | 基于神经网络的目标跟踪方法、装置、设备以及存储介质 |
| CN116369945A (zh) * | 2023-01-30 | 2023-07-04 | 杭州电子科技大学 | 基于4d脉冲神经网络的脑电认知识别方法 |
| CN116407140A (zh) * | 2023-03-13 | 2023-07-11 | 中国科学院深圳先进技术研究院 | 一种脉冲信号检测方法、装置、设备及存储介质 |
| CN116861967A (zh) * | 2022-03-24 | 2023-10-10 | 中国科学院深圳先进技术研究院 | 一种基于脉冲神经网络的数据增强方法及装置 |
Family Cites Families (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101096089B1 (ko) * | 2010-01-28 | 2011-12-19 | 충북대학교 산학협력단 | 수면 무호흡 검출 및 유형 판단 장치 및 그 방법 |
| CN104239694B (zh) * | 2014-08-28 | 2016-11-23 | 北京交通大学 | 一种城轨列车转向架的故障预测与视情维修方法 |
| US10694946B2 (en) * | 2016-07-05 | 2020-06-30 | Freer Logic, Inc. | Dual EEG non-contact monitor with personal EEG monitor for concurrent brain monitoring and communication |
| CN106203628B (zh) * | 2016-07-11 | 2018-12-14 | 深圳先进技术研究院 | 一种增强深度学习算法鲁棒性的优化方法和系统 |
| CN107273688B (zh) * | 2017-06-16 | 2020-03-10 | 山东万腾电子科技有限公司 | 基于原始故障数据的设备寿命与健康度监控方法及系统 |
| JP7083471B2 (ja) * | 2018-09-26 | 2022-06-13 | 株式会社国際電気通信基礎技術研究所 | 推定装置、推定プログラムおよび推定方法 |
| WO2021007615A1 (en) * | 2019-07-12 | 2021-01-21 | Saluda Medical Pty Ltd | Monitoring a quality of neural recordings |
| CN114564814B (zh) * | 2022-01-12 | 2026-04-21 | 云南电网有限责任公司信息中心 | 一种针对稀疏数据的动态阈值高斯核密度估计系统和方法 |
| CN115438312A (zh) * | 2022-10-21 | 2022-12-06 | 国网四川省电力公司成都供电公司 | 一种新能源出力的概率分布模型建立、应用方法及介质 |
| CN116807496B (zh) * | 2023-08-25 | 2023-11-24 | 北京大学 | 癫痫间期脑电波异常信号的定位方法、装置、设备及介质 |
-
2023
- 2023-12-08 CN CN202311675784.3A patent/CN117357134B/zh active Active
- 2023-12-12 WO PCT/CN2023/138053 patent/WO2025118317A1/zh active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106529186A (zh) * | 2016-11-24 | 2017-03-22 | 浙江大学 | 一种脑区脉冲神经信号的预测方法 |
| WO2021249313A1 (zh) * | 2020-06-10 | 2021-12-16 | 上海海事大学 | 一种基于可变分组的偏态分布最优参数估计方法 |
| CN114663468A (zh) * | 2020-12-04 | 2022-06-24 | 株式会社Ntt都科摩 | 基于神经网络的目标跟踪方法、装置、设备以及存储介质 |
| CN114469138A (zh) * | 2022-01-13 | 2022-05-13 | 博睿康医疗科技(上海)有限公司 | 基于时频域的脑电爆发抑制模式的检测方法、系统及介质 |
| CN116861967A (zh) * | 2022-03-24 | 2023-10-10 | 中国科学院深圳先进技术研究院 | 一种基于脉冲神经网络的数据增强方法及装置 |
| CN116369945A (zh) * | 2023-01-30 | 2023-07-04 | 杭州电子科技大学 | 基于4d脉冲神经网络的脑电认知识别方法 |
| CN116407140A (zh) * | 2023-03-13 | 2023-07-11 | 中国科学院深圳先进技术研究院 | 一种脉冲信号检测方法、装置、设备及存储介质 |
Non-Patent Citations (1)
| Title |
|---|
| QI CHUNLIANG : "Automated Image Segmentation Based on Pulse Coupled Neural Networks and Bias-Normal Distribution Index", COMPUTER ENGINEERING AND APPLICATIONS, no. 34, 1 December 2005 (2005-12-01), pages 49 - 51,116, XP093323204 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117357134A (zh) | 2024-01-09 |
| CN117357134B (zh) | 2024-02-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111639535B (zh) | 基于深度学习的人脸识别方法及装置 | |
| JP7414901B2 (ja) | 生体検出モデルのトレーニング方法及び装置、生体検出の方法及び装置、電子機器、記憶媒体、並びにコンピュータプログラム | |
| CN107273845B (zh) | 一种基于置信区域和多特征加权融合的人脸表情识别方法 | |
| CN111986183B (zh) | 一种染色体散型图像自动分割识别系统及装置 | |
| CN110147732A (zh) | 指静脉识别方法、装置、计算机设备及存储介质 | |
| CN114496099B (zh) | 细胞功能注释方法、装置、设备及介质 | |
| WO2024016590A1 (zh) | 一种基于神经拟态计算的在线神经元分类方法 | |
| CN115251953B (zh) | 运动想象脑电信号识别方法、装置、终端设备及存储介质 | |
| CN110672323B (zh) | 一种基于神经网络的轴承健康状态评估方法及装置 | |
| CN114373212A (zh) | 人脸识别模型构建方法、人脸识别方法及相关设备 | |
| WO2025118317A1 (zh) | 一种神经电脉冲检测方法、系统及终端 | |
| CN115512422A (zh) | 基于注意力机制的卷积神经网络面部情绪识别方法及系统 | |
| CN110826624A (zh) | 一种基于深度强化学习的时间序列分类方法 | |
| CN112612844A (zh) | 数据处理方法、装置、设备和存储介质 | |
| CN116206289A (zh) | 一种跨域司机疲劳驾驶检测方法、装置、终端及存储介质 | |
| CN111178387A (zh) | 一种基于多粒度相对密度的标签噪声检测方法 | |
| CN111428151B (zh) | 一种基于网络增速的虚假消息识别方法及其装置 | |
| CN118410332A (zh) | 基于无监督领域自适应的舵机实时故障诊断方法及装置 | |
| CN112925822A (zh) | 基于多表征学习的时间序列分类方法、系统、介质及装置 | |
| CN114631809B (zh) | 一种头部穿戴设备、用眼疲劳监测方法、装置及存储介质 | |
| CN115146693A (zh) | 银行用户的业务流程定制方法及装置 | |
| CN118013308B (zh) | 基于自适应树的时空增强神经电脉冲信号聚类方法及系统 | |
| CN113192629A (zh) | 用于自动胎心判读的方法和装置 | |
| CN119446293A (zh) | 一种单细胞转录组测序数据注释方法 | |
| CN112451857A (zh) | 视网膜神经节细胞激活状态的判别方法、存储介质和设备 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 23960628 Country of ref document: EP Kind code of ref document: A1 |