CN116035576A - Attention mechanism-based depression electroencephalogram signal identification method and system - Google Patents
Attention mechanism-based depression electroencephalogram signal identification method and system Download PDFInfo
- Publication number
- CN116035576A CN116035576A CN202211570940.5A CN202211570940A CN116035576A CN 116035576 A CN116035576 A CN 116035576A CN 202211570940 A CN202211570940 A CN 202211570940A CN 116035576 A CN116035576 A CN 116035576A
- Authority
- CN
- China
- Prior art keywords
- brain
- signal
- electroencephalogram
- original
- signals
- 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
- 238000000034 method Methods 0.000 title claims abstract description 45
- 230000007246 mechanism Effects 0.000 title claims abstract description 25
- 210000004556 brain Anatomy 0.000 claims abstract description 198
- 238000012545 processing Methods 0.000 claims abstract description 45
- 238000013528 artificial neural network Methods 0.000 claims abstract description 27
- 238000009499 grossing Methods 0.000 claims abstract description 21
- 238000007781 pre-processing Methods 0.000 claims abstract description 15
- 230000008569 process Effects 0.000 claims description 20
- 238000011176 pooling Methods 0.000 claims description 11
- 238000006243 chemical reaction Methods 0.000 claims description 10
- 238000005192 partition Methods 0.000 claims description 7
- 238000000605 extraction Methods 0.000 claims description 6
- 238000001914 filtration Methods 0.000 claims description 4
- 230000004044 response Effects 0.000 claims description 4
- 230000003183 myoelectrical effect Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000000513 principal component analysis Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 5
- 238000012549 training Methods 0.000 description 9
- 238000005516 engineering process Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 210000005036 nerve Anatomy 0.000 description 2
- 230000000284 resting effect Effects 0.000 description 2
- 208000014644 Brain disease Diseases 0.000 description 1
- 208000020401 Depressive disease Diseases 0.000 description 1
- 206010054089 Depressive symptom Diseases 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000003759 clinical diagnosis Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000008451 emotion Effects 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000003902 lesion Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000003924 mental process Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 230000002688 persistence Effects 0.000 description 1
- 208000020016 psychiatric disease Diseases 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention belongs to the field of brain-computer interfaces, and particularly relates to a depression brain-electrical signal identification method and system based on an attention mechanism; the method comprises the following steps: acquiring an original brain electrical signal of a tested person and preprocessing the original brain electrical signal to obtain a first brain electrical signal; denoising and smoothing the first electroencephalogram signal to obtain a second electroencephalogram signal; processing the second brain electrical signals by adopting a brain-dividing neural network to obtain a plurality of brain-dividing characteristics; splicing a plurality of brain region features to obtain a whole brain region feature; processing the characteristics of the whole brain region by adopting a neural network of the whole brain region to obtain a recognition result of the brain electrical signal of the depression; the invention uses the attention mechanism to strengthen the weight ratio of different brain region characteristics, has high overall recognition rate and good recognition effect, can assist doctors to diagnose the depression, and has high practicability.
Description
Technical Field
The invention belongs to the field of brain-computer interfaces, and particularly relates to a depression brain-electrical signal identification method and system based on an attention mechanism.
Background
Depression is a mental disorder that primarily affects the mental processes, behavior and emotion, thereby adversely affecting interpersonal relationships and performance. For depression patients who need timely clinical treatment, early accurate diagnosis of depression is critical. For the majority of previous depression diagnoses, questionnaires are used as judgment and screening tools, and this method has the great disadvantage of requiring a doctor with a high experience.
The electroencephalogram signal is a multichannel time sequence and has certain regularity and non-stationarity. It changes with the physiological factors of the individual, and when the brain, especially the cortex, has lesions, the regularity is destroyed and the waveform changes. At present, quantitative measurement is carried out on brain electrical signals obtained from electroencephalogram (EEG), and the method is a nerve imaging technology with obvious practical advantages, and has the advantages of no invasive operation, easy management, good tolerance and relatively low cost. Furthermore, the prevalence and persistence of depressive symptoms makes scalp-recorded electroencephalograms a suitable method for understanding the underlying mechanisms of depression. Therefore, the study of the electroencephalogram can assist the clinical diagnosis of brain diseases.
Considering that EEG network irregularity is one of the physiological symptoms that may be caused by depression, brain electrical activity has spatial features derived from different brain regions, so that brain electrical features extracted from different brain regions can also be used for the identification of depression; the spatial information of the brain can be explored according to the internal structure of the neural network, so that the brain electrical characteristics of different brain regions are obtained, but in the whole brain structure research based on EEG signals, the spatial information of each brain region for exploring depression is rarely researched, and meanwhile, the characteristic weight ratio of the brain regions is a problem to be solved.
Disclosure of Invention
Aiming at the defects existing in the prior art, the invention provides a depression electroencephalogram signal identification method and a depression electroencephalogram signal identification system based on an attention mechanism, wherein the method comprises the following steps:
s1: acquiring an original brain electrical signal of a tested person and preprocessing the original brain electrical signal to obtain a first brain electrical signal;
s2: denoising and smoothing the first electroencephalogram signal to obtain a second electroencephalogram signal;
s3: processing the second brain electrical signals by adopting a brain-dividing neural network to obtain a plurality of brain-dividing characteristics;
s4: splicing a plurality of brain region features to obtain a whole brain region feature;
s5: and processing the characteristics of the whole brain region by adopting the neural network of the whole brain region to obtain the identification result of the brain electrical signal of the depression.
Preferably, the preprocessing of the original electroencephalogram signal includes: selecting brain partition channels of the original brain signals to obtain the original brain signals of a plurality of brain partitions; filtering the original brain electrical signals of a plurality of brain regions by adopting an infinite impulse response digital filter to obtain a first original signal; removing the eye electric signal and the myoelectric signal in the first original signal by adopting independent principal component analysis, and carrying out interpolation processing on the bad channel to obtain a second original signal; and performing downsampling and sliding window processing on the second original signal to obtain a preprocessed original brain electrical signal, namely a first brain electrical signal.
Preferably, the denoising process for the first electroencephalogram signal includes: reconstructing the first electroencephalogram signal into a three-dimensional phase space vector in a phase space; and performing linear least square processing on the three-dimensional phase space vector to obtain a denoised first electroencephalogram signal.
Preferably, the process of smoothing the first electroencephalogram signal includes: performing frequency domain conversion on the denoised first electroencephalogram signal to obtain a frequency domain electroencephalogram signal; and removing outliers on the frequency domain electroencephalogram signals, and performing characteristic smoothing processing on the places from which the outliers are removed to obtain second electroencephalogram signals.
Preferably, the process of processing the second electroencephalogram signal by the brain-partitioned neural network includes: the brain-partitioned neural network comprises a plurality of parallel SENET layers and a convolution layer, wherein the convolution layer comprises a one-dimensional convolution layer, a Relu layer and a maximum pooling layer;
the second electroencephalogram signals are electroencephalogram signals of a plurality of brain regions, the electroencephalogram signals of the brain regions are respectively input into a plurality of parallel SENET layers, and the electroencephalogram signals of the brain regions with the channel weights added are obtained;
and the parallel convolution layer processes the brain electrical signals of the brain regions added with the channel weights to obtain a plurality of brain region characteristics.
Preferably, the process of processing the characteristics of the whole brain region by the whole brain region neural network comprises the following steps: the full brain region neural network comprises a SENET layer, a convolution layer and a full connection layer, wherein the convolution layer comprises a one-dimensional convolution layer, a Relu layer and a maximum pooling layer;
the SENet layer processes the characteristics of the whole brain region to obtain the characteristics of the whole brain region added with the brain region weight;
the whole brain region characteristics added with the brain region weights are sequentially subjected to one-dimensional convolution, a Relu layer and a maximum pooling layer to obtain updated whole brain region characteristics;
and processing the updated brain region characteristics by adopting the full-connection layer to obtain a depression brain electrical signal identification result.
An attention mechanism-based depression electroencephalogram signal recognition system, comprising: the device comprises a data acquisition module, a signal processing module, a signal characteristic extraction module and a signal characteristic classification and identification module;
the data acquisition module is used for acquiring original brain electrical signals of the testee;
the signal processing module is used for carrying out optimization processing on the original electroencephalogram signals of the collected testee to obtain optimized original electroencephalogram signals;
the signal characteristic extraction module is used for extracting characteristics of the optimized original brain signals to obtain the characteristics of the whole brain region;
the signal characteristic classification and identification module is used for carrying out depression brain electrical signal identification according to the characteristics of the whole brain region to obtain a depression brain electrical signal identification result.
Further, the signal processing module comprises a signal preprocessing unit, a time domain signal denoising unit, a time-frequency domain conversion unit and a frequency domain signal smoothing unit;
the signal preprocessing unit is used for preprocessing the original electroencephalogram signals to obtain first electroencephalogram signals;
the time domain signal denoising unit is used for denoising the first electroencephalogram signal to obtain a denoised first electroencephalogram signal;
the time-frequency domain conversion unit is used for converting the denoised first electroencephalogram signal into a frequency domain electroencephalogram signal;
the frequency domain signal smoothing unit is used for carrying out smoothing processing on the frequency domain brain electrical signals to obtain second brain electrical signals, namely optimized brain electrical signals.
The beneficial effects of the invention are as follows: the invention achieves the good denoising effect by using the phase space reconstruction technology on the electroencephalogram signals, processes the electroencephalogram characteristics by using the characteristic smoothing technology on the frequency domain, and effectively avoids the overfitting phenomenon during the training of the neural network by the denoising and smoothing operation; the neural network is combined with the unique structure of the brain, the brain is divided into a plurality of brain areas to be processed by brain electrical signals, and the spatial information and the frequency domain characteristics of channels in the brain areas are effectively extracted; the weight ratio between channels in the brain regions is enhanced by using an attention mechanism, the characteristic relation between the brain regions is considered, the whole characteristic is extracted and identified by using a neural network of the whole brain region, and finally an identification result is obtained; the invention uses the attention mechanism to strengthen the weight ratio of different brain region characteristics, can improve the training parallelism during training, has small training difficulty and high training speed, and simultaneously has high overall recognition rate, good recognition effect and high practicability.
Drawings
FIG. 1 is a flow chart of a method for identifying an electroencephalogram for depression based on an attention mechanism in the invention;
fig. 2 is a schematic diagram of an electroencephalogram feature recognition model structure in the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The invention provides a depression electroencephalogram signal identification method and a depression electroencephalogram signal identification system based on an attention mechanism, as shown in fig. 1, wherein the method comprises the following steps of:
s1: the method comprises the steps of obtaining an original brain electrical signal of a tested person and preprocessing the original brain electrical signal to obtain a first brain electrical signal.
The method for acquiring the original electroencephalogram signal of the whole channel of the testee in the resting state comprises the following steps of: the method comprises the steps of selecting a brain partition channel of an original brain signal of a whole channel, dividing the whole brain region into a plurality of brain regions, and obtaining the original brain signal of the brain regions; preferably, the brain is divided into five brain regions based on the characteristics of the brain and the channel partition of the electroencephalogram cap of the acquisition equipment; because of the compactness between brain regions, the principle of selecting channels is to avoid selecting edge channels between brain regions as much as possible, and the brain region channels are precisely positioned according to the brain position information of each channel in 128-channel HydroCel Geodesic Sensor Net equipment; filtering the original brain electrical signals of a plurality of brain regions by adopting an infinite impulse response digital filter, specifically, carrying out 1-40Hz band-pass filtering on the original brain electrical signals by adopting the infinite impulse response digital filter, and setting the order of the filter to be 6 to obtain a first original signal; removing the eye electric signal and the myoelectric signal in the first original signal by adopting independent principal component analysis, checking the channel, recording the integrity of the signal, and carrying out interpolation processing on the bad channel by using a spherical interpolation method to obtain a second original signal; downsampling and sliding window processing are carried out on the second original signal, specifically, downsampling is carried out on the filtered signal so as to reduce the frequency of the signal to 250Hz; and slicing the sample in a sliding window 2s non-overlapping mode to obtain a preprocessed original brain electrical signal, namely a first brain electrical signal.
S2: and denoising and smoothing the first electroencephalogram signal to obtain a second electroencephalogram signal.
Reconstructing phase space of the electroencephalogram signal by using a coordinate delay method, namely reconstructing a phase space which is the same as that of the original signal in a topological sense by determining delay time t and d-dimensional phase space; for the electroencephalogram signals, the characteristics of the current signals can be determined through the characteristics of the front and rear signals; in the phase space of the signals, if interference noise occurs at a certain moment, the corresponding original brain electrical signals have abrupt change in waveform, so that the brain electrical signals can be judged and corrected in the reconstructed phase space; the denoising process for the first electroencephalogram signal comprises the following steps: reconstructing the first electroencephalogram signal into a three-dimensional phase space vector in a phase space; and performing linear least square processing on the three-dimensional phase space vector, and fitting the current signal point according to the characteristics of the front and rear signals to obtain a denoised first electroencephalogram signal.
The process of smoothing the first electroencephalogram signal comprises the following steps: the denoised first electroencephalogram signal is a time domain signal, and frequency domain conversion is carried out on the denoised first electroencephalogram signal to obtain a frequency domain electroencephalogram signal; and removing the outlier on the frequency domain electroencephalogram through analyzing and judging the outlier of the frequency domain electroencephalogram, and performing characteristic smoothing on the place where the outlier is removed to obtain a second electroencephalogram.
The brain electrical characteristic recognition model comprises a brain region neural network and a whole brain region neural network, and the process of processing the second brain electrical signal by adopting the trained brain electrical characteristic recognition model is as follows:
s3: and processing the second brain electrical signals by adopting a brain-dividing neural network to obtain a plurality of brain-dividing characteristics.
The brain-partitioned neural network comprises a plurality of parallel SENET (compression and excitation network) layers and a convolution layer, wherein the convolution layer comprises a one-dimensional convolution, a Relu layer and a maximum pooling layer, preferably, the convolution kernel of the one-dimensional convolution has a size of 5, the step length is 1, the activation function uses a Relu function, the maximum pooling size is 1*2, and the step length is 2; the SENet layer is an attention mechanism added before the convolution layer, and is sequentially composed of a channel global average pool, a full-connection layer, a Relu layer, a full-connection layer and a sigmoid layer, for example, if the scale of an input electroencephalogram signal is L×40, L is the number of input channels, 40 represents a frequency domain feature point, a 1*L weight is given after the processing of the SENet layer, and the weight is assigned to the processed electroencephalogram signal. As shown in part (a) and (b) of fig. 2, the process of processing the second electroencephalogram signal by using the trained partitioned neural network includes:
the second brain electrical signal is the brain electrical signal of a plurality of brain regions, preferably, the brain region number is 5, namely, the second brain electrical signal of 5 brain regions; the brain electrical signals of a plurality of brain regions are respectively input into a plurality of parallel SENET layers, the SENET layers can adjust the weight of each channel in the brain region signals, and the brain electrical signals of the brain regions with the channel weights added are obtained;
the parallel convolution layer processes the brain electrical signals of the brain regions added with the channel weights to obtain a plurality of brain region division characteristics, wherein the brain region division characteristics comprise characteristic relations among different channels in the brain region.
S4: and splicing the plurality of brain region features to obtain the whole brain region features.
And (3) splicing the characteristics of the brain regions to obtain the characteristics of the whole brain region, and using the characteristics of the whole brain region as the input of the neural network of the whole brain region.
S5: and processing the characteristics of the whole brain region by adopting the neural network of the whole brain region to obtain the identification result of the brain electrical signal of the depression.
As shown in parts (c) (d) (e) of fig. 2, the whole brain region neural network comprises a SENet layer, a convolution layer and a full connection layer, the convolution layer comprises a one-dimensional convolution, a Relu layer and a maximum pooling layer, preferably, the convolution kernel of the one-dimensional convolution has a size of 2, the step size is 1, and Relu is used as an activation layer; the size of the largest pooling layer is 1*2, and the step length is 2; the SENet layer is an attention mechanism added before the convolution layer, and the structure of the SENet layer is the same as that of the SENet layer in the brain-partition nerve network.
The process for processing the characteristics of the whole brain region by adopting the trained neural network for dividing the whole brain region comprises the following steps:
the SENet layer processes the characteristics of the whole brain region, adjusts the weight among the characteristics of each brain region, and obtains the characteristics of the whole brain region after the weight of the brain region is added;
the whole brain region characteristics added with the brain region weights are sequentially subjected to one-dimensional convolution, a Relu layer and a maximum pooling layer, and the relation characteristics among different brain regions are extracted to obtain updated whole brain region characteristics;
and processing the updated whole brain region characteristics by adopting a full-connection layer as a classifier, wherein the number of neurons of the full-connection layer is 2, and finally outputting a result by using a softMax function to finally obtain a depression brain electrical signal identification result (whether the brain electrical signal is a depression brain electrical signal or not).
The invention also provides a depression electroencephalogram signal identification system based on an attention mechanism, which can execute the depression electroencephalogram signal identification method based on the attention mechanism, and comprises the following steps: the device comprises a data acquisition module, a signal processing module, a signal characteristic extraction module and a signal characteristic classification and identification module;
the data acquisition module is used for acquiring original brain electrical signals of the testee; preferably, the data acquisition module comprises an electroencephalogram amplifier, an electroencephalogram cap conforming to the international standard and matched display and storage software. The data acquisition module acquires an electroencephalogram signal of a user in a resting state, performs signal amplification and digital-to-analog conversion, and then transmits the electroencephalogram signal to the PC, wherein the signal transmission adopts a USB communication mode, the electroencephalogram signal acquisition equipment is 128-channel HydroCel Geodesic Sensor Net, the reference electrode is Cz, and the resistance of each electrode channel is kept below 70kΩ.
The signal processing module is used for carrying out optimization processing on the original electroencephalogram signals of the collected testee to obtain optimized original electroencephalogram signals; specific: the signal processing module comprises a signal preprocessing unit, a time domain signal denoising unit, a time-frequency domain conversion unit and a frequency domain signal smoothing unit; the signal preprocessing unit is used for preprocessing the original electroencephalogram signals to obtain first electroencephalogram signals; the time domain signal denoising unit is used for denoising the first electroencephalogram signal to obtain a denoised first electroencephalogram signal; the time-frequency domain conversion unit is used for converting the denoised first electroencephalogram signal into a frequency domain electroencephalogram signal; the frequency domain signal smoothing unit is used for carrying out smoothing processing on the frequency domain electroencephalogram signals to obtain second electroencephalogram signals, namely optimized electroencephalogram signals.
The signal characteristic extraction module is used for extracting characteristics of the optimized original brain signals to obtain the characteristics of the whole brain region;
the signal characteristic classification and identification module is used for carrying out depression brain electrical signal identification according to the characteristics of the whole brain region to obtain a depression brain electrical signal identification result, and the display can be used for displaying the output depression brain electrical signal identification result.
The invention uses the attention mechanism to strengthen the weight ratio of different brain region characteristics, can improve the training parallelism during training, has small training difficulty and high training speed, and simultaneously has high overall recognition rate and good recognition effect, can assist doctors to diagnose depression and has high practicability.
While the foregoing is directed to embodiments, aspects and advantages of the present invention, other and further details of the invention may be had by the foregoing description, it will be understood that the foregoing embodiments are merely exemplary of the invention, and that any changes, substitutions, alterations, etc. which may be made herein without departing from the spirit and principles of the invention.
Claims (8)
1. A depression electroencephalogram signal identification method based on an attention mechanism, which is characterized by comprising the following steps:
s1: acquiring an original brain electrical signal of a tested person and preprocessing the original brain electrical signal to obtain a first brain electrical signal;
s2: denoising and smoothing the first electroencephalogram signal to obtain a second electroencephalogram signal;
s3: processing the second brain electrical signals by adopting a brain-dividing neural network to obtain a plurality of brain-dividing characteristics;
s4: splicing a plurality of brain region features to obtain a whole brain region feature;
s5: and processing the characteristics of the whole brain region by adopting the neural network of the whole brain region to obtain the identification result of the brain electrical signal of the depression.
2. The attention mechanism-based depression electroencephalogram identification method as recited in claim 1, wherein the preprocessing of the original electroencephalogram comprises: selecting brain partition channels of the original brain signals to obtain the original brain signals of a plurality of brain partitions; filtering the original brain electrical signals of a plurality of brain regions by adopting an infinite impulse response digital filter to obtain a first original signal; removing the eye electric signal and the myoelectric signal in the first original signal by adopting independent principal component analysis, and carrying out interpolation processing on the bad channel to obtain a second original signal; and performing downsampling and sliding window processing on the second original signal to obtain a preprocessed original brain electrical signal, namely a first brain electrical signal.
3. The attention mechanism-based depression electroencephalogram identification method as recited in claim 1, wherein the process of denoising the first electroencephalogram comprises: reconstructing the first electroencephalogram signal into a three-dimensional phase space vector in a phase space; and performing linear least square processing on the three-dimensional phase space vector to obtain a denoised first electroencephalogram signal.
4. The attention mechanism-based depression electroencephalogram identification method as recited in claim 1, wherein the process of smoothing the first electroencephalogram comprises: performing frequency domain conversion on the denoised first electroencephalogram signal to obtain a frequency domain electroencephalogram signal; and removing outliers on the frequency domain electroencephalogram signals, and performing characteristic smoothing processing on the places from which the outliers are removed to obtain second electroencephalogram signals.
5. The method for recognizing the brain electrical signals of the depression based on the attention mechanism according to claim 1, wherein the process of processing the second brain electrical signals by the brain region neural network comprises the following steps: the brain-partitioned neural network comprises a plurality of parallel SENET layers and a convolution layer, wherein the convolution layer comprises a one-dimensional convolution layer, a Relu layer and a maximum pooling layer;
the second electroencephalogram signals are electroencephalogram signals of a plurality of brain regions, the electroencephalogram signals of the brain regions are respectively input into a plurality of parallel SENET layers, and the electroencephalogram signals of the brain regions with the channel weights added are obtained;
and the parallel convolution layer processes the brain electrical signals of the brain regions added with the channel weights to obtain a plurality of brain region characteristics.
6. The method for recognizing the brain electrical signals of the depression based on the attention mechanism according to claim 1, wherein the process of processing the characteristics of the whole brain region by the neural network of the whole brain region comprises the following steps: the full brain region neural network comprises a SENET layer, a convolution layer and a full connection layer, wherein the convolution layer comprises a one-dimensional convolution layer, a Relu layer and a maximum pooling layer;
the SENet layer processes the characteristics of the whole brain region to obtain the characteristics of the whole brain region added with the brain region weight;
the whole brain region characteristics added with the brain region weights are sequentially subjected to one-dimensional convolution, a Relu layer and a maximum pooling layer to obtain updated whole brain region characteristics;
and processing the updated brain region characteristics by adopting the full-connection layer to obtain a depression brain electrical signal identification result.
7. An attention mechanism-based depression electroencephalogram signal recognition system, characterized by comprising: the device comprises a data acquisition module, a signal processing module, a signal characteristic extraction module and a signal characteristic classification and identification module;
the data acquisition module is used for acquiring original brain electrical signals of the testee;
the signal processing module is used for carrying out optimization processing on the original electroencephalogram signals of the collected testee to obtain optimized original electroencephalogram signals;
the signal characteristic extraction module is used for extracting characteristics of the optimized original brain signals to obtain the characteristics of the whole brain region;
the signal characteristic classification and identification module is used for carrying out depression brain electrical signal identification according to the characteristics of the whole brain region to obtain a depression brain electrical signal identification result.
8. The attention mechanism-based depression electroencephalogram signal recognition system as recited in claim 7, wherein the signal processing module comprises a signal preprocessing unit, a time domain signal denoising unit, a time-frequency domain conversion unit and a frequency domain signal smoothing unit;
the signal preprocessing unit is used for preprocessing the original electroencephalogram signals to obtain first electroencephalogram signals;
the time domain signal denoising unit is used for denoising the first electroencephalogram signal to obtain a denoised first electroencephalogram signal;
the time-frequency domain conversion unit is used for converting the denoised first electroencephalogram signal into a frequency domain electroencephalogram signal;
the frequency domain signal smoothing unit is used for carrying out smoothing processing on the frequency domain brain electrical signals to obtain second brain electrical signals, namely optimized brain electrical signals.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211570940.5A CN116035576A (en) | 2022-12-08 | 2022-12-08 | Attention mechanism-based depression electroencephalogram signal identification method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211570940.5A CN116035576A (en) | 2022-12-08 | 2022-12-08 | Attention mechanism-based depression electroencephalogram signal identification method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116035576A true CN116035576A (en) | 2023-05-02 |
Family
ID=86119029
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211570940.5A Pending CN116035576A (en) | 2022-12-08 | 2022-12-08 | Attention mechanism-based depression electroencephalogram signal identification method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116035576A (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113012816A (en) * | 2021-04-12 | 2021-06-22 | 东北大学 | Brain partition risk prediction method and device, electronic equipment and storage medium |
| CN113367679A (en) * | 2021-07-05 | 2021-09-10 | 北京银河方圆科技有限公司 | Target point determination method, device, equipment and storage medium |
| CN113397563A (en) * | 2021-07-22 | 2021-09-17 | 北京脑陆科技有限公司 | Training method, device, terminal and medium for depression classification model |
| CN113907756A (en) * | 2021-09-18 | 2022-01-11 | 深圳大学 | Wearable system of physiological data based on multiple modalities |
| CN114021608A (en) * | 2021-11-17 | 2022-02-08 | 南京工业大学 | Electroencephalogram recognition method fusing structure information between brain regions |
| CN114504331A (en) * | 2022-02-25 | 2022-05-17 | 北京工业大学 | Mood recognition and classification method fusing CNN and LSTM |
| CN114521903A (en) * | 2022-02-15 | 2022-05-24 | 南京邮电大学 | Electroencephalogram attention recognition system and method based on feature selection |
| CN115204231A (en) * | 2022-07-19 | 2022-10-18 | 北京交通大学 | Digital human-computer interface cognitive load assessment method based on EEG (electroencephalogram) multi-dimensional features |
-
2022
- 2022-12-08 CN CN202211570940.5A patent/CN116035576A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113012816A (en) * | 2021-04-12 | 2021-06-22 | 东北大学 | Brain partition risk prediction method and device, electronic equipment and storage medium |
| CN113367679A (en) * | 2021-07-05 | 2021-09-10 | 北京银河方圆科技有限公司 | Target point determination method, device, equipment and storage medium |
| CN113397563A (en) * | 2021-07-22 | 2021-09-17 | 北京脑陆科技有限公司 | Training method, device, terminal and medium for depression classification model |
| CN113907756A (en) * | 2021-09-18 | 2022-01-11 | 深圳大学 | Wearable system of physiological data based on multiple modalities |
| CN114021608A (en) * | 2021-11-17 | 2022-02-08 | 南京工业大学 | Electroencephalogram recognition method fusing structure information between brain regions |
| CN114521903A (en) * | 2022-02-15 | 2022-05-24 | 南京邮电大学 | Electroencephalogram attention recognition system and method based on feature selection |
| CN114504331A (en) * | 2022-02-25 | 2022-05-17 | 北京工业大学 | Mood recognition and classification method fusing CNN and LSTM |
| CN115204231A (en) * | 2022-07-19 | 2022-10-18 | 北京交通大学 | Digital human-computer interface cognitive load assessment method based on EEG (electroencephalogram) multi-dimensional features |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP7018133B2 (en) | ECG heartbeat automatic identification classification method based on artificial intelligence | |
| CN114266276B (en) | Motor imagery electroencephalogram signal classification method based on channel attention and multi-scale time domain convolution | |
| EP3692903A1 (en) | Artificial intelligence self-learning-based automatic electrocardiography analysis method and apparatus | |
| CN109645980A (en) | A kind of rhythm abnormality classification method based on depth migration study | |
| CN110069958A (en) | A kind of EEG signals method for quickly identifying of dense depth convolutional neural networks | |
| CN111166327A (en) | Epilepsy diagnosis device based on single-channel electroencephalogram signal and convolutional neural network | |
| CN112120694A (en) | Motor imagery electroencephalogram signal classification method based on neural network | |
| CN109567789B (en) | Electrocardiogram data segmentation processing method and device and readable storage medium | |
| CN112450947B (en) | Dynamic brain network analysis method for emotional arousal degree | |
| CN111428601B (en) | P300 signal identification method, device and storage medium based on MS-CNN | |
| CN112674782B (en) | Device and method for detecting epileptic-like electrical activity of epileptic during inter-seizure period | |
| CN111202512A (en) | Electrocardiogram classification method and device based on wavelet transformation and DCNN | |
| KR20190073330A (en) | Method and apparatus for an automatic artifact removal of EEG based on a deep leaning algorithm | |
| CN111931578B (en) | Electroencephalogram identification method based on minimum spanning tree and regional double-layer network | |
| CN114190944A (en) | Robust emotion recognition method based on electroencephalogram signals | |
| CN118094317A (en) | Motor imagery electroencephalogram signal classification system based on TimesNet and convolutional neural network | |
| CN115017960B (en) | Electroencephalogram signal classification method based on space-time combined MLP network and application | |
| CN116421200A (en) | Brain electricity emotion analysis method of multi-task mixed model based on parallel training | |
| CN116035576A (en) | Attention mechanism-based depression electroencephalogram signal identification method and system | |
| CN115736920A (en) | Depression state identification method and system based on bimodal fusion | |
| Zhang et al. | An atrial fibrillation classification method based on an outlier data filtering strategy and modified residual block of the feature pyramid network | |
| Tihak et al. | Classification of Atrial Fibrillation ECG Signals Using 2D CNN | |
| CN117860201B (en) | Feature classification method and system based on deep learning analysis model | |
| Li et al. | Efficient ECG classification method for arrhythmia using MODWPT and adaptive incremental broad learning | |
| Subudhi et al. | DeEN: Deep Ensemble Framework for Neuroatypicality Classification |
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 |