CN112932501A - Method for automatically identifying insomnia based on one-dimensional convolutional neural network - Google Patents
Method for automatically identifying insomnia based on one-dimensional convolutional neural network Download PDFInfo
- Publication number
- CN112932501A CN112932501A CN202110094678.0A CN202110094678A CN112932501A CN 112932501 A CN112932501 A CN 112932501A CN 202110094678 A CN202110094678 A CN 202110094678A CN 112932501 A CN112932501 A CN 112932501A
- Authority
- CN
- China
- Prior art keywords
- neural network
- convolutional neural
- layer
- insomnia
- dimensional
- 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
- 208000013738 Sleep Initiation and Maintenance disease Diseases 0.000 title claims abstract description 51
- 206010022437 insomnia Diseases 0.000 title claims abstract description 51
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 45
- 238000000034 method Methods 0.000 title claims abstract description 38
- 230000008667 sleep stage Effects 0.000 claims abstract description 21
- 238000012549 training Methods 0.000 claims abstract description 17
- 238000012360 testing method Methods 0.000 claims abstract description 12
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 238000003062 neural network model Methods 0.000 claims abstract description 4
- 239000013598 vector Substances 0.000 claims description 25
- 238000011176 pooling Methods 0.000 claims description 18
- 230000004913 activation Effects 0.000 claims description 11
- 238000010606 normalization Methods 0.000 claims description 6
- 239000000284 extract Substances 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 5
- 230000009467 reduction Effects 0.000 claims description 4
- 238000006073 displacement reaction Methods 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 239000000126 substance Substances 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 abstract description 4
- 239000010410 layer Substances 0.000 description 57
- 230000006870 function Effects 0.000 description 17
- 239000008186 active pharmaceutical agent Substances 0.000 description 8
- 238000000605 extraction Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 2
- 238000003745 diagnosis Methods 0.000 description 2
- 201000010099 disease Diseases 0.000 description 2
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000003595 spectral effect Effects 0.000 description 2
- 208000024172 Cardiovascular disease Diseases 0.000 description 1
- 206010020772 Hypertension Diseases 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 206010012601 diabetes mellitus Diseases 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 238000013399 early diagnosis Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000004630 mental health Effects 0.000 description 1
- 230000037053 non-rapid eye movement Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000001766 physiological effect Effects 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 238000013138 pruning Methods 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 230000004461 rapid eye movement Effects 0.000 description 1
- 238000013077 scoring method Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 239000002356 single layer Substances 0.000 description 1
- 208000019116 sleep disease Diseases 0.000 description 1
- 208000020685 sleep-wake disease Diseases 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/48—Other medical applications
- A61B5/4806—Sleep evaluation
- A61B5/4809—Sleep detection, i.e. determining whether a subject is asleep or not
-
- 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
- A61B5/7225—Details of analog processing, e.g. isolation amplifier, gain or sensitivity adjustment, filtering, baseline or drift compensation
-
- 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
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
Abstract
The invention discloses a method for automatically identifying insomnia based on a one-dimensional convolutional neural network, which comprises the following steps: preprocessing an original single-channel electroencephalogram signal data set to remove high-frequency noise and direct-current components in the electroencephalogram signals; constructing a subdata set from the preprocessed electroencephalogram signals according to sleep stages and an overlapping method of different periods; constructing a one-dimensional convolution neural network model based on the reconstructed subdata set; and inputting the test set signal into the trained one-dimensional convolutional neural network for classification and identification. By applying the embodiment of the invention, the step of manually extracting the features is removed, the limitation of prior knowledge is eliminated, the features related to insomnia are automatically extracted through the training and learning of the neural network, the information related to insomnia in the sequence can be fully extracted, deep coding is carried out, the complexity of the traditional model is solved, the workload is greatly reduced, the efficiency of automatically identifying the insomnia is effectively improved, and the accuracy is improved.
Description
Technical Field
The invention relates to the technical field of biomedicine, in particular to a method for automatically identifying insomnia based on a one-dimensional convolutional neural network.
Background
Sleep is a basic physiological activity and plays an important role in the physical and mental health of human bodies. Insomnia is a typical sleep disorder commonly existing in the general population, which does not threaten the life safety of people in a short period of time, but if left to develop, causes a series of concurrent diseases to patients and even leads to death. The early diagnosis and treatment of insomnia can reduce the risk of diseases such as depression, diabetes, hypertension, cardiovascular diseases and the like, reduce a series of troubles caused by insomnia to life and improve the life quality of people.
Diagnosis of insomnia is typically based on a sleep questionnaire on the patient, polysomnography data, and insomnia diagnostic criteria published by the american society for sleep medicine. The insomnia severity index is a valid clinical indicator that uses a short self-reporting tool to quantify a patient's perception of insomnia. However, in the process of insomnia diagnosis, the traditional manual method lacks effective physiological indicators for automatically identifying insomnia, is time-consuming and labor-consuming, and is easily influenced by the subjective of doctors. Therefore, insomnia cannot be diagnosed efficiently and accurately by the conventional method.
At present, various sleep stage automatic scoring methods based on electroencephalogram signals have high identification performance for insomnia, but most of the methods are traditional classifiers which need to be trained by manually extracting sleep features based on experience. The Aydin and the like extract 10-dimensional singular spectral characteristics of the sleep electroencephalogram signals and send the signals into a single-layer artificial neural network for insomnia identification. Hamida et al extracts spectral and Hjorth parameter characteristics and applies principal component analysis to reduce dimension. The first principal component is then classified according to a set threshold. Zhang et al proposed a method for insomnia identification based on time, spectrum, nonlinear features and random forest classifiers. Therefore, most of the existing methods for automatically identifying insomnia are based on manually extracted features and traditional machine learning algorithms, are limited by a priori knowledge, and take much time and energy. The convolutional neural network does not need to define the characteristics manually, local and global characteristics of the input signal can be learned through training, the problem of insufficient characteristic extraction caused by manually extracting the characteristics is solved, the limitation that the manually extracted characteristics are limited by priori knowledge is eliminated, and the insomnia automatic identification efficiency is greatly improved. Therefore, the convolutional neural network is applied to the electroencephalogram signal, namely the field of spatially discrete unstructured data processing, so that the classification performance of the electroencephalogram signal insomnia identification is improved.
Disclosure of Invention
The invention aims to provide a method for automatically identifying insomnia based on a one-dimensional convolutional neural network, which aims to solve the problem of insufficient feature extraction caused by the existing manual feature extraction, simultaneously get rid of the limitation that the manual feature extraction is limited by priori knowledge, and improve the efficiency of automatically identifying insomnia.
In order to achieve the purpose, the invention provides a method for automatically identifying insomnia based on a one-dimensional convolutional neural network, which comprises the following steps:
preprocessing an original single-channel electroencephalogram signal data set to remove high-frequency noise and direct-current components in the electroencephalogram signals;
constructing a subdata set from the preprocessed electroencephalogram signals according to sleep stages and an overlapping method of different periods;
constructing a one-dimensional convolution neural network model based on the reconstructed subdata set;
and inputting the test set signal into the trained one-dimensional convolutional neural network for classification and identification.
Optionally, the step of constructing the sub-data set from the preprocessed electroencephalogram signals according to sleep stages and overlapping methods of different periods includes:
selecting a sleep stage of a subdata set required to be constructed;
selecting two successive period time window signal segments;
if the selected time window signal is the same as the selected sleep stage, overlapping by using a sliding window to construct a subdata set; otherwise, no overlap is performed.
In one implementation, the step of inputting the test set signal to the trained one-dimensional convolutional neural network for classification and identification includes:
inputting the test set signal into the convolutional layer to fully extract the characteristics related to insomnia in the sequence;
using the pooling layers to reduce the dimension of the features, and mapping the features into one-dimensional vectors in the last pooling layer;
and sending all the one-dimensional vectors into a full connection layer for feature fusion, and inputting the vectors into a softmax classifier for classification to obtain a classification result.
Optionally, the step of constructing a one-dimensional convolutional neural network model based on the reconstructed sub data set includes:
selecting a convolutional neural network, wherein the convolutional neural network comprises: 5 convolutional layers, 3 pooling layers and 3 full-link layers; the method comprises the following steps that a first convolution layer and a second convolution layer respectively use 1 x 11 and 1 x 5 large convolution kernels, a third convolution layer, a fourth convolution layer and a fifth convolution layer use 1 x 3 small convolution kernels, after the first convolution layer, the second convolution layer and the fifth convolution layer, the features are subjected to dimensionality reduction by using 1 x 3 pooling layers, the features are mapped into one-dimensional vectors in the last pooling layer, and the vectors are sent to a full-connection layer for secondary classification;
and training the selected convolutional neural network based on the reconstructed subdata set, and taking the trained convolutional neural network as a one-dimensional convolutional neural network model.
In one implementation, the step of training the convolutional neural network based on the reconstructed sub-data set includes: inputting a one-dimensional time sequence of 30s epoch of a designated sleep stage to the selected convolutional neural network, wherein the selected convolutional neural network uses a one-dimensional convolutional kernel which is specifically expressed as:
wherein
For the ith pixel of the ith layer output feature,
and blRepresenting the weight vector and deviation parameter of the convolution kernel in the l-th layer, d is the size of the convolution kernel, and N is the input characteristic vector
Length of (2)And f (-) represents the activation function of the convolutional layer.
After the first convolution layer, the second convolution layer and the fifth convolution layer, dimensionality reduction is performed on the features by using the pooling layers with the size of 1 x 3, the features are mapped into one-dimensional vectors through the last pooling layer, the vectors are sent to the full-link layer to be subjected to secondary classification, and a final recognition result is obtained, wherein ReLU is used as an activation function and is specifically expressed as follows:
f(x)＝max(0,x)
wherein x is a feature matrix output after convolution;
the first convolution layer is followed by a batch normalization layer that normalizes the feature map before entering the activation function to reduce the internal covariate bias, the batch normalization being defined as follows:
wherein B represents a small batch consisting of m samples,. mu.BAnd
respectively representing the mean and variance of B, with ε being a constant and γ and β being calculated during the training
Scale and displacement parameters;
and the full connection layer is classified by using a SoftMax classifier to complete classification output of the insomnia signals.
Optionally, the classification function of the SoftMax classifier is specifically expressed as:
wherein
The output value of the ith node is shown, and m represents the number of classes.
The method for automatically identifying insomnia based on the one-dimensional convolutional neural network provided by the embodiment of the invention has the following beneficial effects:
(1) according to the invention, electroencephalogram feature extraction and classification are integrated into an algorithm, the step of manually extracting features is removed, the limitation of priori knowledge is eliminated, the features related to insomnia are automatically extracted through training and learning of a neural network, information related to insomnia in a sequence can be fully extracted, deep coding is carried out, the complexity of a traditional model is solved, the workload is greatly reduced, the efficiency of automatically identifying insomnia is effectively improved, and the accuracy is improved.
(2) In the existing research, it is not clear which electroencephalogram sleep stages have the best insomnia identification performance, the method disclosed by the invention can be used for automatically identifying insomnia by constructing the subdata set, so that the classification accuracy is improved, and meanwhile, the performance of automatically identifying insomnia based on different sleep stages can be further obtained by comparison, so that the effective sleep stage for identifying insomnia is determined, and the experience is provided for further researching insomnia later.
Drawings
FIG. 1 is a schematic flow chart of a method for automatically identifying insomnia based on a one-dimensional convolutional neural network according to an embodiment of the present invention.
Fig. 2 is a schematic diagram of a method for overlapping signal segments for two time windows of consecutive periods.
FIG. 3 is a schematic diagram of constructing a sub data set.
FIG. 4 is a model diagram of a one-dimensional convolutional neural network according to an embodiment of the present invention.
Detailed Description
The embodiments of the present invention are described below with reference to specific embodiments, and other advantages and effects of the present invention will be easily understood by those skilled in the art from the disclosure of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention.
Please refer to fig. 1-4. It should be noted that the drawings provided in the present embodiment are only for illustrating the basic idea of the present invention, and the components related to the present invention are only shown in the drawings rather than drawn according to the number, shape and size of the components in actual implementation, and the type, quantity and proportion of the components in actual implementation may be changed freely, and the layout of the components may be more complicated.
As shown in fig. 1, the present invention provides a method for automatically identifying insomnia based on one-dimensional convolutional neural network, comprising:
s110, preprocessing an original single-channel electroencephalogram signal data set to remove high-frequency noise and direct-current components in the electroencephalogram signal;
it should be noted that, because EEG is a low-frequency signal, and the frequency components of the EEG are mainly concentrated in the frequency band of 0.5-50Hz, the invention designs an 80-order FIR band-pass filter (0.5-50Hz) based on a hamming window to perform filtering preprocessing on the original single-channel electroencephalogram signal, so as to remove high-frequency noise and direct-current components.
S120, constructing a subdata set from the preprocessed electroencephalogram signals according to sleep stages and overlapping methods of different periods;
it is understood that the sleep stages of the sub data sets to be constructed are selected, the sub data sets are constructed according to an overlap method, two continuous period time window signal segments are selected, if the same sleep stages exist, sliding window overlap is used, and if different sleep stages exist, overlap is not performed. Specifically, according to the R & K rule, a night electroencephalogram sleep signal is divided into epochs of 30 seconds, each epoch is divided into an awake phase (Wake), a rapid eye movement phase (REM) and a non-rapid eye movement phase (NREM), wherein the NREM is divided into stages S1, S2, S3 and S4, S1 and S2 are combined into a sub data set LSS-DS, S3 and S4 are combined into a sub data set SWS-DS, and the sub data set ALL-DS and the REM phase sub data set REM-DS comprise ALL the stages. Thus, four subdata sets, ALL-DS, REM-DS, LSS-DS and SWS-DS, are constructed according to different epochs of sleep stages and an overlap method, as shown in fig. 2-3.
S130, constructing a one-dimensional convolution neural network model based on the reconstructed subdata set;
it should be noted that the convolutional neural network of the present invention includes 11 layers, which are five convolutional layers, three pooling layers and three fully-connected layers, respectively, and fig. 4 is a model diagram of a one-dimensional convolutional neural network. The input to the model is a one-dimensional time series of specified sleep stages 30s epoch, so the design model uses a one-dimensional convolution kernel. The one-dimensional convolution operation process is defined as:
wherein the content of the first and second substances,
for the ith pixel of the ith layer output feature,
and blRepresenting the weight vector and deviation parameter of the convolution kernel in the l-th layer, d is the size of the convolution kernel, and N is the input characteristic vector
F (-) represents the activation function of the convolutional layer.
The first convolutional layer and the second convolutional layer use large convolution kernels of 1 × 11 and 1 × 5, respectively, and the third convolutional layer, the fourth convolutional layer, and the fifth convolutional layer use small convolution kernels of 1 × 3. After the first, second, and fifth convolutional layers, the features are reduced in dimension using pooling layers of size 1 × 3, and the last pooling layer maps the features to one-dimensional vectors. And sending the vector into a full connection layer for secondary classification to obtain a final identification result. At this module, the ReLU (Whole line Unit) is chosen as the activation function, which is defined as follows:
f(x)＝max(0,x)
wherein x is a feature matrix output after convolution;
the first convolution layer is followed by a batch normalization layer that normalizes the feature map before entering the activation function, thereby reducing internal covariate shifts. Batch normalization is defined as follows:
wherein B represents a small batch consisting of m samples,. mu.BAnd
mean and variance of B are expressed, respectively, with epsilon being a constant and gamma and beta being scale and displacement parameters calculated during the training process.
And the full connection layer is classified by using a SoftMax classifier to complete classification output of the insomnia signals.
Since the single fully-connected layer tends to have non-linearity, the present invention uses three fully-connected layers to solve the possible non-linearity problem. The full link layer also adopts a linear ReLU function as an activation function to avoid the problem of gradient explosion. And finally, performing final insomnia classification prediction by using a SoftMax function.
Wherein the content of the first and second substances,
the output value of the ith node is shown, and m represents the number of classes.
And S140, inputting the test set signal to the trained one-dimensional convolutional neural network for classification and identification.
It will be appreciated that the sleep signals of all epochs are mixed together and then divided into a training set, a validation set and a test set according to a specified ratio. Inputting the processed EEG signals into a convolutional neural network, iteratively updating network parameters to enable the classification performance of the model to be optimal, and then testing and classifying the test set input model to obtain a sleep signal classification result.
Specifically, a training set signal is input to carry out iterative optimization training on the model, an AdaMod optimizer and a cross entropy loss function are adopted to optimize network parameters, and a ReLU activation function is adopted to reduce the interdependence relationship among the parameters.
The method comprises the steps of initializing weight parameters in a one-dimensional convolutional neural network by using a kaiming initializer, improving the convergence rate of a model, optimizing the parameters by using an AdaMod optimizer, calculating the index long-term average value of the adaptive learning rate during training, and pruning the overhigh learning rate during the training by using the average value, so that the sensitivity to the learning rate is reduced, and the convergence is improved. The invention employs a cross entropy loss function, which is defined as:
where N is the number of training samples. y iskAnd pkRepresenting the true and predicted phases of the sample, respectively.
In order to prevent the over-fitting problem, Dropout and an L2 regularization method are used, wherein Dropout randomly breaks off units in a layer with a designated probability in the training process, and L2 regularization is to add a regularization term after a loss function to reduce the complexity of a network. The cross-entropy loss function has a function definition of the L2 regularization term as:
wherein E isnIs a basic cross-entropy loss function that,
is an L2 regular term, and λ and w are penalty factors and network parameters, respectively.
And finally, inputting the test set signals into a one-dimensional convolutional neural network, fully extracting features related to insomnia in the sequence by the convolutional layer, reducing the dimensions of the features by using pooling layers, mapping the features into one-dimensional vectors in the last pooling layer, sending all the one-dimensional vectors into a full connection layer for feature fusion, and inputting the vectors into a softmax classifier for classification to obtain a classification result.
The invention integrates the feature extraction and classification into an algorithm, removes the step of manually extracting features, effectively eliminates the limitation of prior knowledge, can fully extract information related to insomnia in a sequence by automatically extracting the features through a neural network, carries out deep coding, solves the problem of complexity of a traditional model, greatly reduces the workload, improves the intelligence, enables the extraction and the processing of electroencephalogram signals to be more intelligent and convenient, effectively improves the efficiency of automatically identifying the insomnia, gets rid of the limitation of the prior knowledge, enlarges the research space of automatically identifying the insomnia, and improves the accuracy.
The foregoing embodiments are merely illustrative of the principles and utilities of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or change the above-mentioned embodiments without departing from the spirit and scope of the present invention. Accordingly, it is intended that all equivalent modifications or changes which can be made by those skilled in the art without departing from the spirit and technical spirit of the present invention be covered by the claims of the present invention.
Claims (6)
1. A method for automatically identifying insomnia based on a one-dimensional convolutional neural network is characterized by comprising the following steps:
preprocessing an original single-channel electroencephalogram signal data set to remove high-frequency noise and direct-current components in the electroencephalogram signals;
constructing a subdata set from the preprocessed electroencephalogram signals according to sleep stages and an overlapping method of different periods;
constructing a one-dimensional convolution neural network model based on the reconstructed subdata set;
and inputting the test set signal into the trained one-dimensional convolutional neural network for classification and identification.
2. The method for automatically identifying insomnia based on one-dimensional convolutional neural network as claimed in claim 1, wherein said step of constructing the sub data set from the preprocessed electroencephalogram signal according to sleep stages and overlapping methods of different periods:
selecting a sleep stage of a subdata set required to be constructed;
selecting two successive period time window signal segments;
if the selected time window signal is the same as the selected sleep stage, overlapping by using a sliding window to construct a subdata set; otherwise, no overlap is performed.
3. The method for automatically identifying insomnia based on one-dimensional convolutional neural network as claimed in claim 1, wherein said step of inputting test set signal to trained one-dimensional convolutional neural network for classification and identification:
inputting the test set signal into the convolutional layer to fully extract the characteristics related to insomnia in the sequence;
reducing the dimension of the features by using a pooling layer, and mapping the features into one-dimensional vectors;
and sending all the one-dimensional vectors into a full connection layer for feature fusion, and inputting the vectors into a softmax classifier for classification to obtain a classification result.
4. The method of claim 1, wherein the step of constructing a one-dimensional convolutional neural network model based on the reconstructed sub data sets comprises:
selecting a convolutional neural network, wherein the convolutional neural network comprises: 5 convolutional layers, 3 pooling layers and 3 full-link layers; the method comprises the following steps that a first convolution layer and a second convolution layer respectively use 1 x 11 and 1 x 5 large convolution kernels, a third convolution layer, a fourth convolution layer and a fifth convolution layer use 1 x 3 small convolution kernels, after the first convolution layer, the second convolution layer and the fifth convolution layer, the features are subjected to dimensionality reduction by using 1 x 3 pooling layers, the features are mapped into one-dimensional vectors in the last pooling layer, and the vectors are sent to a full-connection layer for secondary classification;
and training the convolutional neural network based on the reconstructed subdata set, and taking the trained convolutional neural network as a one-dimensional convolutional neural network model.
5. The method of claim 4, wherein the step of training the convolutional neural network based on the reconstructed subdata set comprises: for the one-dimensional time sequence of sleep stages 30s epoch specified by the input of the selected convolutional neural network, the selected convolutional neural network uses a one-dimensional convolutional kernel which is specifically expressed as:
wherein the content of the first and second substances,
for the ith pixel of the ith layer output feature,
and blRepresenting the weight vector and deviation parameter of the convolution kernel in the l-th layer, d is the size of the convolution kernel, and N is the input characteristic vector
F (-) represents the activation function of the convolutional layer;
after the first convolution layer, the second convolution layer and the fifth convolution layer, dimensionality reduction is performed on the features by using a pooling layer with the size of 1 x 3, the features are mapped into one-dimensional vectors through the pooling layer, the vectors are sent into a full-connection layer to be subjected to secondary classification, a final identification result is obtained, and the ReLU is used as an activation function and is specifically expressed as follows:
f(x)＝max(0,x)
wherein x is a feature matrix output after convolution;
the first convolution layer is followed by a batch normalization layer that normalizes features before inputting into the activation function to reduce internal covariate shifts, the batch normalization being defined as follows:
wherein B represents a small batch consisting of m samples,. mu.BAnd
respectively representing the mean and variance of B, epsilon is a constant, and gamma and beta are scale and displacement parameters calculated in the training process;
and the full connection layer is classified by using a SoftMax classifier to complete classification output of the insomnia signals.
6. The method for automatically identifying insomnia based on one-dimensional convolutional neural network as claimed in claim 5, wherein the classification function of said SoftMax classifier is specifically expressed as:
wherein
The output value of the ith node is shown, and m represents the number of classes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110094678.0A CN112932501A (en) | 2021-01-25 | 2021-01-25 | Method for automatically identifying insomnia based on one-dimensional convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110094678.0A CN112932501A (en) | 2021-01-25 | 2021-01-25 | Method for automatically identifying insomnia based on one-dimensional convolutional neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112932501A true CN112932501A (en) | 2021-06-11 |
Family
ID=76236256
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110094678.0A Pending CN112932501A (en) | 2021-01-25 | 2021-01-25 | Method for automatically identifying insomnia based on one-dimensional convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112932501A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113962231A (en) * | 2021-10-13 | 2022-01-21 | 杭州胜铭纸业有限公司 | Optical identification comparison method and system for information codes of packing cases |
| CN114098763A (en) * | 2021-12-13 | 2022-03-01 | 清华大学深圳国际研究生院 | Electroencephalogram denoising method |
| CN115429293A (en) * | 2022-11-04 | 2022-12-06 | 之江实验室 | Sleep type classification method and device based on impulse neural network |
Citations (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060019224A1 (en) * | 2004-07-23 | 2006-01-26 | Pics, Inc. | Insomnia assessment and treatment device and method |
| DE69637337D1 (en) * | 1995-11-08 | 2008-01-10 | Oxford Biosignals Ltd | Improvements in terms of physiological monitoring |
| US20110190594A1 (en) * | 2010-02-04 | 2011-08-04 | Robert Bosch Gmbh | Device and method to monitor, assess and improve quality of sleep |
| KR101480536B1 (en) * | 2014-02-27 | 2015-01-12 | 주식회사 올비트앤 | Electroencephalogram detecting system including a portable electroencephalogram detecting apparatus of hair band type, and a sleep management method using the same |
| KR101498812B1 (en) * | 2013-09-06 | 2015-03-05 | 세종대학교산학협력단 | Insomnia tests and derived indicators using eeg |
| US20150190086A1 (en) * | 2014-01-03 | 2015-07-09 | Vital Connect, Inc. | Automated sleep staging using wearable sensors |
| US20170360362A1 (en) * | 2014-12-05 | 2017-12-21 | Agency For Science, Technology And Research | Sleep profiling system with feature generation and auto-mapping |
| CN108310587A (en) * | 2018-02-02 | 2018-07-24 | 贺鹏程 | A kind of sleep control device and method |
| CN108969863A (en) * | 2018-05-23 | 2018-12-11 | 上海海事大学 | assisting sleep and progressive wake-up system |
| CN110163180A (en) * | 2019-05-29 | 2019-08-23 | 长春思帕德科技有限公司 | Mental imagery eeg data classification method and system |
| CN110327040A (en) * | 2019-04-24 | 2019-10-15 | 武汉理工大学 | Sleep stage method and system based on cloud platform |
| CN110584596A (en) * | 2019-07-15 | 2019-12-20 | 天津大学 | Sleep stage classification method based on dual-input convolutional neural network and application |
| CN110623665A (en) * | 2019-09-26 | 2019-12-31 | 川北医学院 | Intelligent sleep time phase detection and sleep quality evaluation system and method |
| US20200054289A1 (en) * | 2016-02-01 | 2020-02-20 | Verily Life Sciences Llc | Machine learnt model to detect rem sleep periods using a spectral analysis of heart rate and motion |
| CN110897639A (en) * | 2020-01-02 | 2020-03-24 | 清华大学深圳国际研究生院 | Electroencephalogram sleep staging method based on deep convolutional neural network |
| WO2020085553A1 (en) * | 2018-10-25 | 2020-04-30 | 고려대학교 산학협력단 | Apparatus and method for inducing sleep by using neurofeedback |
| CN111387936A (en) * | 2019-01-02 | 2020-07-10 | 中国移动通信有限公司研究院 | Sleep stage identification method, device and equipment |
| CN111783534A (en) * | 2020-05-28 | 2020-10-16 | 东南大学 | Sleep staging method based on deep learning |
-
2021
- 2021-01-25 CN CN202110094678.0A patent/CN112932501A/en active Pending
Patent Citations (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE69637337D1 (en) * | 1995-11-08 | 2008-01-10 | Oxford Biosignals Ltd | Improvements in terms of physiological monitoring |
| US20060019224A1 (en) * | 2004-07-23 | 2006-01-26 | Pics, Inc. | Insomnia assessment and treatment device and method |
| US20110190594A1 (en) * | 2010-02-04 | 2011-08-04 | Robert Bosch Gmbh | Device and method to monitor, assess and improve quality of sleep |
| KR101498812B1 (en) * | 2013-09-06 | 2015-03-05 | 세종대학교산학협력단 | Insomnia tests and derived indicators using eeg |
| US20150190086A1 (en) * | 2014-01-03 | 2015-07-09 | Vital Connect, Inc. | Automated sleep staging using wearable sensors |
| KR101480536B1 (en) * | 2014-02-27 | 2015-01-12 | 주식회사 올비트앤 | Electroencephalogram detecting system including a portable electroencephalogram detecting apparatus of hair band type, and a sleep management method using the same |
| US20170360362A1 (en) * | 2014-12-05 | 2017-12-21 | Agency For Science, Technology And Research | Sleep profiling system with feature generation and auto-mapping |
| US20200054289A1 (en) * | 2016-02-01 | 2020-02-20 | Verily Life Sciences Llc | Machine learnt model to detect rem sleep periods using a spectral analysis of heart rate and motion |
| CN108310587A (en) * | 2018-02-02 | 2018-07-24 | 贺鹏程 | A kind of sleep control device and method |
| CN108969863A (en) * | 2018-05-23 | 2018-12-11 | 上海海事大学 | assisting sleep and progressive wake-up system |
| WO2020085553A1 (en) * | 2018-10-25 | 2020-04-30 | 고려대학교 산학협력단 | Apparatus and method for inducing sleep by using neurofeedback |
| CN111387936A (en) * | 2019-01-02 | 2020-07-10 | 中国移动通信有限公司研究院 | Sleep stage identification method, device and equipment |
| CN110327040A (en) * | 2019-04-24 | 2019-10-15 | 武汉理工大学 | Sleep stage method and system based on cloud platform |
| CN110163180A (en) * | 2019-05-29 | 2019-08-23 | 长春思帕德科技有限公司 | Mental imagery eeg data classification method and system |
| CN110584596A (en) * | 2019-07-15 | 2019-12-20 | 天津大学 | Sleep stage classification method based on dual-input convolutional neural network and application |
| CN110623665A (en) * | 2019-09-26 | 2019-12-31 | 川北医学院 | Intelligent sleep time phase detection and sleep quality evaluation system and method |
| CN110897639A (en) * | 2020-01-02 | 2020-03-24 | 清华大学深圳国际研究生院 | Electroencephalogram sleep staging method based on deep convolutional neural network |
| CN111783534A (en) * | 2020-05-28 | 2020-10-16 | 东南大学 | Sleep staging method based on deep learning |
Non-Patent Citations (2)
| Title |
|---|
| ZHU T, 等: "Convolution-and attention-based neural network for automated sleep stage classification", INTERNATIONAL JOURNAL OF ENVIRONMENTAL RESEARCH AND PUBLIC HEALTH * |
| 谢宏，等: "基于离散小波变换的脑电信号睡眠分期研究", 软件与算法 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113962231A (en) * | 2021-10-13 | 2022-01-21 | 杭州胜铭纸业有限公司 | Optical identification comparison method and system for information codes of packing cases |
| CN113962231B (en) * | 2021-10-13 | 2024-03-26 | 杭州胜铭纸业有限公司 | Packaging box information code optical identification comparison method and system |
| CN114098763A (en) * | 2021-12-13 | 2022-03-01 | 清华大学深圳国际研究生院 | Electroencephalogram denoising method |
| CN115429293A (en) * | 2022-11-04 | 2022-12-06 | 之江实验室 | Sleep type classification method and device based on impulse neural network |
| CN115429293B (en) * | 2022-11-04 | 2023-04-07 | 之江实验室 | Sleep type classification method and device based on impulse neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109389059B (en) | P300 detection method based on CNN-LSTM network | |
| CN112932501A (en) | Method for automatically identifying insomnia based on one-dimensional convolutional neural network | |
| CN114052735B (en) | Deep field self-adaption-based electroencephalogram emotion recognition method and system | |
| CN110897639A (en) | Electroencephalogram sleep staging method based on deep convolutional neural network | |
| CN111493828B (en) | Sequence-to-sequence sleep disorder detection method based on full convolution network | |
| CN104636580A (en) | Health monitoring mobile phone based on human face | |
| Liu et al. | Research on medical data feature extraction and intelligent recognition technology based on convolutional neural network | |
| CN114203295B (en) | Cerebral apoplexy risk prediction intervention method and system | |
| CN114366124B (en) | Epileptic electroencephalogram identification method based on semi-supervised deep convolution channel attention list classification network | |
| CN112641451B (en) | Multi-scale residual error network sleep staging method and system based on single-channel electroencephalogram signal | |
| CN108567418A (en) | A kind of pulse signal inferior health detection method and detecting system based on PCANet | |
| CN111028232A (en) | Diabetes classification method and equipment based on fundus images | |
| CN114595725B (en) | Electroencephalogram signal classification method based on addition network and supervised contrast learning | |
| CN113705398A (en) | Music electroencephalogram space-time characteristic classification method based on convolution-long and short term memory network | |
| Barhate et al. | Analysis of classifiers for prediction of type ii diabetes mellitus | |
| CN116058800A (en) | Automatic sleep stage system based on deep neural network and brain-computer interface | |
| CN111047590A (en) | Hypertension classification method and device based on fundus images | |
| Das et al. | Automated classification of retinal OCT images using a deep multi-scale fusion CNN | |
| CN114300126A (en) | Cancer prediction system based on early cancer screening questionnaire and feed-forward neural network | |
| CN113925459A (en) | Sleep staging method based on electroencephalogram feature fusion | |
| CN113974655A (en) | Epileptic seizure prediction method based on electroencephalogram signals | |
| CN113421250A (en) | Intelligent fundus disease diagnosis method based on lesion-free image training | |
| Qu et al. | Epileptogenic region detection based on deep CNN with transfer learning | |
| Begawan et al. | Sleep stage identification based on EEG signals using parallel convolutional neural network and recurrent neural network | |
| CN115399735A (en) | Multi-head attention mechanism sleep staging method based on time-frequency double-current enhancement |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |