AU2020100027A4 - Electroencephalogram-based negative emotion recognition method and system for aggressive behavior prediction - Google Patents
Electroencephalogram-based negative emotion recognition method and system for aggressive behavior prediction Download PDFInfo
- Publication number
- AU2020100027A4 AU2020100027A4 AU2020100027A AU2020100027A AU2020100027A4 AU 2020100027 A4 AU2020100027 A4 AU 2020100027A4 AU 2020100027 A AU2020100027 A AU 2020100027A AU 2020100027 A AU2020100027 A AU 2020100027A AU 2020100027 A4 AU2020100027 A4 AU 2020100027A4
- Authority
- AU
- Australia
- Prior art keywords
- feature vector
- negative emotion
- initial
- emotion
- sample
- 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.)
- Ceased
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Public Health (AREA)
- Data Mining & Analysis (AREA)
- Biomedical Technology (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Artificial Intelligence (AREA)
- Primary Health Care (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Epidemiology (AREA)
- Databases & Information Systems (AREA)
- Pathology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Signal Processing (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The present invention discloses an electroencephalogram (EEG)-based negative emotion recognition method and system for aggressive behavior prediction. The method includes: conducting processing and feature extraction on obtained sample data to obtain an initial emotion sample feature vector, where the sample data includes EEG signals generated by stimulating healthy subjects in multiple negative emotion stimulation modes and a negative emotion stimulation mode corresponding to each EEG signal; training a deep neural network based on the initial emotion sample feature vector, and determining a middle-layer feature of a trained deep neural network model as an optimized sample feature vector; training a classifier according to the optimized sample feature vector and the initial emotion sample feature vector to determine a negative emotion recognition and classification model; and processing an EEG signal of a subject, and recognizing a negative emotion of the subject according to a processed EEG signal of the subject and the negative emotion recognition and classification model. The present invention can increase an EEG-based classification and recognition rate of emotions, so as to avoid and prevent an aggressive behavior. Obtain sample data 101 Preprocess EEG signals in the sample data Conduct feature extraction on preprocessed sample data to 103 obtain an initial emotion sample feature vector Train a deep neural network based on the initial emotion sample feature vector and a negative emotion stimulation mode corresponding to the initial emotion sample feature, to obtain a 104 trained deep neural network model, and determine a middle layer feature of the trained deep neural network model as an optimized sample feature vector Train a classifier according tothe optimized sample feature vector and the initial emotion sample feature vector to 105 determine a negative emotion recognition and classification model Obtain and process an EEG signal of a subject Recognize a negative emotion of the subject according to a 107 processed EEG signal of the subject and the negative emotion recognition and classification model
Description
ELECTROENCEPHALOGRAM-BASED NEGATIVE EMOTION RECOGNITION METHOD AND SYSTEM FOR AGGRESSIVE BEHAVIOR PREDICTION TECHNICAL FIELD
The present invention relates to the field of electroencephalogram (EEG)-based negative emotion recognition technologies, and in particular, to an EEG-based negative emotion recognition method and system for aggressive behavior prediction.
BACKGROUND
In recent years, violence incidents among college students occurred frequently. As a young and impulsive group, college students are more susceptible to negative emotions and thus exhibit aggressive behaviors, antithetical behaviors, disruptive behaviors, or other violent behaviors. These aggressive behaviors are closely related to negative emotions. A strong negative emotion with high arousal is highly destructive. The aggressive behavior can be regarded as an external manifestation of this emotional state.
SUMMARY
An objective of the present invention is to provide an EEG-based negative emotion recognition method and system for aggressive behavior prediction, which can increase an EEG-based classification and recognition rate of emotions, so as to avoid and prevent an aggressive behavior.
To achieve the above purpose, the present invention provides the following technical solution:
An EEG-based negative emotion recognition method for aggressive behavior prediction includes:
obtaining sample data, where the sample data includes multiple data groups; each data group includes EEG signals generated by stimulating healthy subjects of a same sex and age in multiple negative emotion stimulation modes and a negative emotion stimulation mode corresponding to each EEG signal; and different data groups correspond to different ages or sexes of healthy subjects;
preprocessing the EEG signals in the sample data;
conducting feature extraction on preprocessed sample data to obtain an initial emotion sample feature vector;
training a deep neural network based on the initial emotion sample feature vector and a negative emotion stimulation mode corresponding to the initial emotion sample feature vector, to obtain a trained deep neural network model; and determining a middle-layer feature of the trained deep neural network model as an optimized sample feature vector;
training a classifier according to the optimized sample feature vector and the initial emotion
2020100027 08 Jan 2020 sample feature vector to determine a negative emotion recognition and classification model;
obtaining and processing an EEG signal of a subject; and recognizing a negative emotion of the subject according to a processed EEG signal of the subject and the negative emotion recognition and classification model.
An EEG-based negative emotion recognition system for aggressive behavior prediction includes:
a sample data obtaining module, configured to obtain sample data, where the sample data includes multiple data groups; each data group includes EEG signals generated by stimulating healthy subjects of a same sex and age in multiple negative emotion stimulation modes and a negative emotion stimulation mode corresponding to each EEG signal; and different data groups correspond to different ages or sexes of healthy subjects;
a preprocessing module, configured to preprocess the EEG signals in the sample data;
an initial emotion sample feature vector obtaining module, configured to conduct feature extraction on preprocessed sample data to obtain an initial emotion sample feature vector;
an optimized sample feature vector determining module, configured to train a deep neural network based on the initial emotion sample feature vector and a negative emotion stimulation mode corresponding to the initial emotion sample feature vector, to obtain a trained deep neural network model; and determine a middle-layer feature of the trained deep neural network model as an optimized sample feature vector;
a negative emotion recognition and classification model determining module, configured to train a classifier according to the optimized sample feature vector and the initial emotion sample feature vector to determine a negative emotion recognition and classification model;
an EEG signal obtaining and processing module, configured to obtain and process an EEG signal of a subject; and a negative emotion recognition module, configured to recognize a negative emotion of the subject according to a processed EEG signal of the subject and the negative emotion recognition and classification model.
According to specific embodiments provided in the present invention, the present invention discloses the following technical effects.
According to the EEG-based negative emotion recognition method and system for aggressive behavior prediction, a negative emotion recognition and classification model is determined according to a large amount of sample data and a deep learning technology; and a negative emotion of a subject is recognized by using a processed EEG signal of the subject in combination with a negative emotion recognition and classification model. The present invention can increase an EEG-based classification and recognition rate of emotions, so as to avoid and prevent an
2020100027 08 Jan 2020 aggressive behavior.
BRIEF DESCRIPTION OF THE DRAWINGS
To describe the technical solutions in the embodiments of the present invention or in the prior art more clearly, the following briefly introduces the accompanying drawings required for describing the embodiments. Apparently, the accompanying drawings in the following description show merely some embodiments of the present invention, and a person of ordinary skill in the art may still derive other accompanying drawings from these accompanying drawings without creative efforts.
FIG. 1 is a schematic flowchart of an EEG-based negative emotion recognition method for aggressive behavior prediction according to an embodiment of the present invention;
FIG. 2 is a flowchart of determining a negative emotion recognition and classification model according to an embodiment of the present invention;
FIG. 3 is a frame diagram of conducting emotional feature extraction in real time based on a deep neural network according to an embodiment of the present invention; and
FIG. 4 is a schematic structural diagram of an EEG-based negative emotion recognition system for aggressive behavior prediction according to an embodiment of the present invention; DETAILED DESCRIPTION
The following clearly and completely describes the technical solutions in the embodiments of the present invention with reference to accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are merely a part rather than all of the embodiments of the present invention. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.
An aggressive behavior is closely related to a negative emotion. To some extent, the cognition of a negative emotion with high arousal can be used as a basis for predicting and assessing the occurrence of an aggressive behavior and crime. The present invention proposes an EEG-based negative emotion recognition method and system for aggressive behavior prediction. Cognitive analysis and classified prediction are conducted on a negative emotion with relatively high arousal, to determine and assess a psychological state of a subject. Corresponding emotion regulation and psychotherapy measures are taken for a predicted subject that is prone to generate a highly negative emotion, so as to avoid and prevent an aggressive behavior.
In order to make the above objects, features, and advantages of the present invention more apparent, the present invention will be further described in detail in connection with the accompanying drawings and the detailed description.
FIG. 1 is a schematic flowchart of an EEG-based negative emotion recognition method for
2020100027 08 Jan 2020 aggressive behavior prediction according to an embodiment of the present invention. FIG. 2 is a flowchart of determining a negative emotion recognition and classification model according to an embodiment of the present invention. With reference to FIG. 1 and FIG. 2, the EEG-based negative emotion recognition method for aggressive behavior prediction provided in the present invention includes the following steps.
Step 101. Obtain sample data, where the sample data includes multiple data groups; each data group includes EEG signals generated by stimulating healthy subjects of a same sex and age in multiple negative emotion stimulation modes and a negative emotion stimulation mode corresponding to each EEG signal; and different data groups correspond to different ages or sexes of healthy subjects.
Step 101 specifically includes determining ages and sexes of healthy subjects and conducting grouping experiments on the healthy subjects according to the ages and sexes of the healthy subjects.
Multiple negative emotion stimulation modes are used to stimulate each of these grouped healthy subjects; EEG signals of these healthy subjects are acquired; and the EEG signals and a negative emotion stimulation mode corresponding to each EEG signal form sample data. In the present invention, a multichannel signal acquisition mode is used.
Step 102. Preprocess the EEG signals in the sample data.
An EEG signal is a type of very weak signal. Therefore, in an acquisition process, an EEG signal is prone to be interfered by another noise signal. Preprocessing an EEG signals mainly means removing all artifacts in the acquired EEG signal. These artifacts mainly include an electro-oculogram (EOG) signal, an EMG signal, an electrocardio signal, a power line interference signal, an electromagnetic interference signal, and a task-irrelevant EEG signal. An EEG signal after artifact removal is used for feature extraction for different analysis and research. To reduce the impact of an artifact on EEG signal analysis, it is necessary to conduct filtering and de-noising processing on an EEG signal.
In the present invention, an isolated component analysis (ICA) algorithm and an original data waveform feature are used to conduct de-noising on the EEG signals (the sample data), to obtain preprocessed sample data. The original data waveform feature signal herein generally refers to a specific waveform feature in the EEG signal, such as an EOG signal or an electrocardio signal. A corresponding independent component (IC) signal waveform in a sample data signal calculated by using the ICA algorithm is found according to waveform features such as an EOG signal and a heartbeat signal. Then, the IC signal waveform is removed to complete noise signal removal.
Step 103. Conduct feature extraction on preprocessed sample data to obtain an initial emotion sample feature vector.
2020100027 08 Jan 2020
According to a time point at which stimulation is conducted by using a first negative image, EEG signals in various negative emotion stimulation stages are successively extracted to construct an epoch set. For a power spectrum of the epoch set of the whole EEG data, time corresponding to a same negative emotion stimulation mode is locked; and an average power change of the power spectrum is calculated according to frequencies, so as to obtain a two-dimensional time-frequency spectrum of event-related spectral perturbation (ERSP). Feature analysis is conducted on the two-dimensional time-frequency spectrum of the event-related spectral perturbation (ERSP). A frequency range and brain region distribution with a high recognition rate are obtained through determining according to a one-way analysis of variance algorithm (one-way ANOVA). Then, corresponding power spectrum density (PSD) values are extracted for the frequency range and brain region distribution with a high recognition rate to construct a high-dimensional feature vector as the initial emotion sample feature vector.
Step 104. Train a deep neural network based on the initial emotion sample feature vector and a negative emotion stimulation mode corresponding to the initial emotion sample feature vector, to obtain a trained deep neural network model; and determine a middle-layer feature of the trained deep neural network model as an optimized sample feature vector.
The deep neural network model is used for deep feature extraction. The deep neural network can directly associate a feature with a category to which the feature belongs, that is, use a multi-layer network structure of the deep neural network to implement mapping from a feature at an input layer to a category to which the feature belongs. According to the idea of deep learning, output of nodes at each layer can be regarded as projections of a same representation in different feature spaces. Obviously, a middle-layer feature acquired from the middle of the network has a specific distinguishing feature.
In the deep neural network (DNN), a space of an original feature is projected to a new feature space formed by various nodes at a hidden layer, to obtain a feature representation form of the space of an original feature at a first hidden layer. Finally, the feature is mapped to a state space at an output layer through a Softmax network, and is associated with a category to which the feature belongs, as shown in FIG. 3. A hidden layer in the middle of the network is finally selected through nonlinear mapping of multiple hidden layers by using original input of the network, and node output of the hidden layer is acquired as a new feature. A key of this feature extraction algorithm is how to train a deep learning model by using original emotion PSD features.
Gaussian nodes are selected for modeling at an input layer and the last hidden layer (corresponding to a middle feature output layer) of the network. Bernoulli nodes are selected for modeling at other hidden layers. Softmax nodes are selected for modeling at the output layer.
Input and output of Softmax nodes satisfy the following:
2020100027 08 Jan 2020 _ exp(x) y' Σ,6χρ(χ/)
Xi and J, respectively represent input and output of the nodes.
A training algorithm of the deep neural network is as follows:
First, a deep belief network (DBN) layer-by-layer pre-training method is used to initialize network parameters. In a network parameter tuning stage, an emotional state id corresponding to each frame of training data is obtained through forced alignment, and is used as a training label; and training data is used as a stimulation signal and is mapped to the output layer through various hidden layers, and a node whose output value is the largest is recorded as a predicted label. A minimum cross entropy is used as an objective function, and a parameter is adjusted by using an error back propagation algorithm. The objective function is recorded as follows:
FXENT(^) = EZi^G)10g /(0 t(0
Θ represents a network parameter; n represents a total number of different emotion labels; j(z) represents an output value of a node z in the network output layer when training data is given, that is, a predicted occurrence probability of an emotional state z; and j'(z) represents an actual occurrence probability of the emotional state z.
After training of the deep neural network is completed, the network parameter of the deep neural network is used to map original input to obtain a new feature of the emotion, that is, y=f(x, Qdnn) x represents original acoustic feature input; Odnn represents a deep neural network parameter; and j represents a middle-layer feature extracted after mapping by the deep neural network.
Step 105. Train a classifier according to the optimized sample feature vector and the initial emotion sample feature vector to determine a negative emotion recognition and classification model.
An optimized characteristic parameter extracted based on the deep learning model and a PSD characteristic parameter obtained based on ERSP analysis are respectively introduced to commonly used and popular classifier models, for example, a support vector machine algorithm, a K-nearest neighbor algorithm, a linear discriminant analysis algorithm, a Naive Bayes algorithm, and a random forest algorithm. Negative emotion recognition accuracy rates of the models are compared for analysis to determine a negative emotion recognition and classification model suitable for high arousal. In the foregoing steps, multiple classifier algorithms are selected. In the present invention, a combination of different algorithms (feature extraction model +
2020100027 08 Jan 2020 classification model) is selected, to obtain a classification model with an optimal combination scheme. Specific operation steps include: training a classifier according to the optimized sample feature vector and the initial emotion sample feature vector to determine multiple initial negative emotion recognition and classification models; and conducting screening on the initial negative emotion recognition and classification models according to a cross validation algorithm by using a minimum error rate in cross-validation as a standard, to determine a final negative emotion recognition and classification model.
Step 106. Obtain and process an EEG signal of a subject. A processing procedure of step 106 is the same as that of step 102 to step 104.
Step 107. Recognize a negative emotion of the subject according to a processed EEG signal of the subject and the negative emotion recognition and classification model.
To achieve the foregoing objective, the present invention further provides an EEG-based negative emotion recognition system for aggressive behavior prediction. As shown in FIG. 4, the system includes the following modules:
a sample data obtaining module 201, configured to obtain sample data, where the sample data includes multiple data groups; each data group includes EEG signals generated by stimulating healthy subjects of a same sex and age in multiple negative emotion stimulation modes and a negative emotion stimulation mode corresponding to each EEG signal; and different data groups correspond to different ages or sexes of healthy subjects;
a preprocessing module 202, configured to preprocess the EEG signals in the sample data;
an initial emotion sample feature vector obtaining module 203, configured to conduct feature extraction on preprocessed sample data to obtain an initial emotion sample feature vector;
an optimized sample feature vector determining module 204, configured to train a deep neural network based on the initial emotion sample feature vector and a negative emotion stimulation mode corresponding to the initial emotion sample feature vector, to obtain a trained deep neural network model; and determine a middle-layer feature of the trained deep neural network model as an optimized sample feature vector;
a negative emotion recognition and classification model determining module 205, configured to train a classifier according to the optimized sample feature vector and the initial emotion sample feature vector to determine a negative emotion recognition and classification model;
an EEG signal obtaining and processing module 206, configured to obtain and process an EEG signal of a subject; and a negative emotion recognition module 207, configured to recognize a negative emotion of the subject according to a processed EEG signal of the subject and the negative emotion recognition and classification model.
2020100027 08 Jan 2020
The preprocessing module 202 specifically includes:
a preprocessing unit, configured to conduct de-noising processing on the EEG signals by using an isolated component analysis algorithm and an original data waveform feature, to obtain the preprocessed sample data.
The initial emotion sample feature vector obtaining module 203 specifically includes:
an epoch set construction unit, configured to: for each data group, according to a time point at which stimulation is conducted by using a first negative emotion stimulation mode successively extract preprocessed EEG signals in the data group that are corresponding to the negative emotion stimulation modes, and form an epoch set by using the EEG signals;
a two-dimensional time-frequency spectrum determining unit, configured to: for a power spectrum of the epoch set, lock time corresponding to a same negative emotion stimulation mode, and conduct calculation according to frequencies to obtain a two-dimensional time-frequency spectrum of event-related spectral perturbation;
a frequency range and brain region distribution determining unit, configured to conduct feature analysis on the two-dimensional time-frequency spectrum, and obtain a frequency range and brain region distribution through determining according to a one-way analysis of variance algorithm; and an initial emotion sample feature vector obtaining unit, configured to extract PSD values according to the frequency range and the brain region distribution, and construct a high-dimensional feature vector by using the PSD values, as the initial emotion sample feature vector.
The negative emotion recognition and classification model determining module 205 specifically includes:
an initial negative emotion recognition and classification model determining unit, configured to train the classifier according to the optimized sample feature vector and the initial emotion sample feature vector to determine multiple initial negative emotion recognition and classification models; and a negative emotion recognition and classification model determining unit, configured to conduct screening on the initial negative emotion recognition and classification models according to a cross validation algorithm by using a minimum error rate in cross-validation as a standard, to determine a final negative emotion recognition and classification model.
To prevent aggressive crimes, especially aggressive behaviors of a college student group, in the present invention, a deep learning model is used for classification and recognition to propose and establish an EEG-based aggressive behavior prediction model during the research on negative emotion cognition and aggressive behavior prediction. Cognitive analysis and classified
2020100027 08 Jan 2020 prediction are conducted on a negative emotion with relatively high arousal, to determine and assesse a psychological state of a subject. Corresponding emotion regulation and psychotherapy measures are taken for a predicted subject that is prone to generate a highly negative emotion, so as to avoid and prevent an aggressive behavior. The research content and achievement further enrich cognition of the prediction model of an EEG-based aggressive behavior tendency, deepen understanding of a human brain information processing mechanism in a negative emotion elicitation process. In addition, the research content and achievement improve flexibility, efficiency, and practicability of a current negative emotion recognition mode, and also greatly improve accuracy of the mode, thereby accelerating the actual application process of an EEG-based model for prediction of tendency of an aggressive behavior resulting from a negative emotion. In addition, analysis and summarization of EEG signals generated by stimulation of negative emotions also provide a reference basis for the cognitive physiology field.
Each embodiment of the present specification is described in a progressive manner, each embodiment focuses on the difference from other embodiments, and the same and similar parts between the embodiments may refer to each other. For a system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and reference can be made to the method description.
Several examples are used for illustration of the principles and implementation methods of the present invention. The description of the embodiments is used to help illustrate the method and its core principles of the present invention. In addition, a person of ordinary skill in the art can make various modifications in terms of specific embodiments and scope of application in accordance with the teachings of the present invention. In conclusion, the content of this specification shall not be construed as a limitation to the present invention.
Claims (5)
1. An electroencephalogram (EEG)-based negative emotion recognition method for aggressive behavior prediction, comprising:
obtaining sample data, wherein the sample data comprises multiple data groups; each data group comprises EEG signals generated by stimulating healthy subjects of a same sex and age in multiple negative emotion stimulation modes and a negative emotion stimulation mode corresponding to each EEG signal; and different data groups correspond to different ages or sexes of healthy subjects;
preprocessing the EEG signals in the sample data;
conducting feature extraction on preprocessed sample data to obtain an initial emotion sample feature vector;
training a deep neural network based on the initial emotion sample feature vector and a negative emotion stimulation mode corresponding to the initial emotion sample feature vector, to obtain a trained deep neural network model; and determining a middle-layer feature of the trained deep neural network model as an optimized sample feature vector;
training a classifier according to the optimized sample feature vector and the initial emotion sample feature vector to determine a negative emotion recognition and classification model;
obtaining and processing an EEG signal of a subject; and recognizing a negative emotion of the subject according to a processed EEG signal of the subject and the negative emotion recognition and classification model.
2. The EEG-based negative emotion recognition method for aggressive behavior prediction according to claim 1, wherein the preprocessing the EEG signals in the sample data specifically comprises conducting de-noising processing on the EEG signals by using an isolated component analysis algorithm and an original data waveform feature, to obtain preprocessed sample data;
wherein the conducting feature extraction on preprocessed sample data to obtain an initial emotion sample feature vector specifically comprises:
for each data group, according to a time point at which stimulation is conducted by using a first negative emotion stimulation mode, successively extracting preprocessed EEG signals in the data group that are corresponding to the negative emotion stimulation modes, and constructing an epoch set by using the EEG signals;
for a power spectrum of the epoch set, locking time corresponding to a same negative emotion stimulation mode, and conducting calculation according to frequencies to obtain a two-dimensional time-frequency spectrum of event-related spectral perturbation;
conducting feature analysis on the two-dimensional time-frequency spectrum, and obtaining
2020100027 08 Jan 2020 a frequency range and brain region distribution through determining according to a one-way analysis of variance algorithm; and extracting PSD values according to the frequency range and the brain region distribution, and constructing a high-dimensional feature vector by using the PSD values, as the initial emotion sample feature vector;
wherein nodes at a network input layer and the last hidden layer of the trained deep neural network model are Gaussian nodes; nodes at other hidden layers are Bernoulli nodes; and a node at an output layer is a Softmax node;
wherein the training a classifier according to the optimized sample feature vector and the initial emotion sample feature vector to determine a negative emotion recognition and classification model specifically comprises:
training the classifier according to the optimized sample feature vector and the initial emotion sample feature vector to determine multiple initial negative emotion recognition and classification models; and conducting screening on the initial negative emotion recognition and classification models according to a cross validation algorithm by using a minimum error rate in cross-validation as a standard, to determine a final negative emotion recognition and classification model.
3. An EEG-based negative emotion recognition system for aggressive behavior prediction, comprising:
a sample data obtaining module, configured to obtain sample data, wherein the sample data comprises multiple data groups; each data group comprises EEG signals generated by stimulating healthy subjects of a same sex and age in multiple negative emotion stimulation modes and a negative emotion stimulation mode corresponding to each EEG signal; and different data groups correspond to different ages or sexes of healthy subjects;
a preprocessing module, configured to preprocess the EEG signals in the sample data;
an initial emotion sample feature vector obtaining module, configured to conduct feature extraction on preprocessed sample data to obtain an initial emotion sample feature vector;
an optimized sample feature vector determining module, configured to train a deep neural network based on the initial emotion sample feature vector and a negative emotion stimulation mode corresponding to the initial emotion sample feature vector, to obtain a trained deep neural network model; and determine a middle-layer feature of the trained deep neural network model as an optimized sample feature vector;
a negative emotion recognition and classification model determining module, configured to train a classifier according to the optimized sample feature vector and the initial emotion sample feature vector to determine a negative emotion recognition and classification model;
2020100027 08 Jan 2020 an EEG signal obtaining and processing module, configured to obtain and process an EEG signal of a subject; and a negative emotion recognition module, configured to recognize a negative emotion of the subject according to a processed EEG signal of the subject and the negative emotion recognition and classification model.
4. The EEG-based negative emotion recognition system for aggressive behavior prediction according to claim 6, wherein the preprocessing module specifically comprises:
a preprocessing unit, configured to conduct de-noising processing on the EEG signals by using an isolated component analysis algorithm and an original data waveform feature, to obtain preprocessed sample data.
5. The EEG-based negative emotion recognition system for aggressive behavior prediction according to claim 6, wherein the initial emotion sample feature vector obtaining module specifically comprises:
an epoch set construction unit, configured to: for each data group, according to a time point at which stimulation is conducted by using a first negative emotion stimulation mode, successively extract preprocessed EEG signals in the data group that are corresponding to the negative emotion stimulation modes, and construct an epoch set by using the EEG signals;
a two-dimensional time-frequency spectrum determining unit, configured to: for a power spectrum of the epoch set, lock time corresponding to a same negative emotion stimulation mode, and conduct calculation according to frequencies to obtain a two-dimensional time-frequency spectrum of event-related spectral perturbation;
a frequency range and brain region distribution determining unit, configured to conduct feature analysis on the two-dimensional time-frequency spectrum, and obtain a frequency range and brain region distribution through determining according to a one-way analysis of variance algorithm; and an initial emotion sample feature vector obtaining unit, configured to extract PSD values according to the frequency range and the brain region distribution, and construct a high-dimensional feature vector by using the PSD values, as the initial emotion sample feature vector;
wherein the negative emotion recognition and classification model determining module specifically comprises:
an initial negative emotion recognition and classification model determining unit, configured to train the classifier according to the optimized sample feature vector and the initial emotion sample feature vector to determine multiple initial negative emotion recognition and classification models; and a negative emotion recognition and classification model determining unit, configured to conduct screening on the initial negative emotion recognition and classification models according to a cross validation algorithm by using a minimum error rate in cross-validation as a standard, to determine a final negative emotion recognition and classification model.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911141842.8A CN110826527A (en) | 2019-11-20 | 2019-11-20 | Electroencephalogram negative emotion recognition method and system based on aggressive behavior prediction |
| CN201911141842.8 | 2019-11-20 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| AU2020100027A4 true AU2020100027A4 (en) | 2020-02-20 |
Family
ID=69526426
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| AU2020100027A Ceased AU2020100027A4 (en) | 2019-11-20 | 2020-01-08 | Electroencephalogram-based negative emotion recognition method and system for aggressive behavior prediction |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN110826527A (en) |
| AU (1) | AU2020100027A4 (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111860463A (en) * | 2020-08-07 | 2020-10-30 | 北京师范大学 | Emotion identification method based on joint norm |
| CN113723365A (en) * | 2021-09-29 | 2021-11-30 | 西安电子科技大学 | Target feature extraction and classification method based on millimeter wave radar point cloud data |
| CN114662541A (en) * | 2022-03-24 | 2022-06-24 | 西人马(深圳)科技有限责任公司 | Fault diagnosis model construction method and device and electronic equipment |
| CN114997315A (en) * | 2022-06-13 | 2022-09-02 | 西安交通大学 | Error correlation potential classification method based on multichannel electroencephalogram integration |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111579939A (en) * | 2020-04-23 | 2020-08-25 | 天津大学 | Method for detecting partial discharge phenomenon of high-voltage power cable based on deep learning |
| CN111557660A (en) * | 2020-06-08 | 2020-08-21 | 东北大学 | Arrhythmia identification method under sub-population deep learning framework |
| CN112163518B (en) * | 2020-09-28 | 2023-07-18 | 华南理工大学 | Emotion modeling method for emotion monitoring and adjusting system |
| CN113128353B (en) * | 2021-03-26 | 2023-10-24 | 安徽大学 | Emotion perception method and system oriented to natural man-machine interaction |
| CN113191225B (en) * | 2021-04-19 | 2023-07-04 | 华南师范大学 | Emotion electroencephalogram recognition method and system based on graph attention network |
| CN113180669B (en) * | 2021-05-12 | 2024-04-26 | 中国人民解放军中部战区总医院 | Emotion adjustment training system and method based on nerve feedback technology |
| CN113317802A (en) * | 2021-05-21 | 2021-08-31 | 上海六河创意设计有限公司 | Information early warning system and method based on brain waves |
| CN113486752B (en) * | 2021-06-29 | 2023-06-16 | 吉林大学 | Emotion recognition method and system based on electrocardiosignal |
-
2019
- 2019-11-20 CN CN201911141842.8A patent/CN110826527A/en not_active Withdrawn
-
2020
- 2020-01-08 AU AU2020100027A patent/AU2020100027A4/en not_active Ceased
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111860463A (en) * | 2020-08-07 | 2020-10-30 | 北京师范大学 | Emotion identification method based on joint norm |
| CN111860463B (en) * | 2020-08-07 | 2024-02-02 | 北京师范大学 | Emotion recognition method based on joint norm |
| CN113723365A (en) * | 2021-09-29 | 2021-11-30 | 西安电子科技大学 | Target feature extraction and classification method based on millimeter wave radar point cloud data |
| CN113723365B (en) * | 2021-09-29 | 2024-05-14 | 西安电子科技大学 | Millimeter wave Lei Dadian cloud data-based target feature extraction and classification method |
| CN114662541A (en) * | 2022-03-24 | 2022-06-24 | 西人马(深圳)科技有限责任公司 | Fault diagnosis model construction method and device and electronic equipment |
| CN114997315A (en) * | 2022-06-13 | 2022-09-02 | 西安交通大学 | Error correlation potential classification method based on multichannel electroencephalogram integration |
| CN114997315B (en) * | 2022-06-13 | 2024-07-16 | 西安交通大学 | Multi-channel electroencephalogram integration-based error related potential classification method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110826527A (en) | 2020-02-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| AU2020100027A4 (en) | Electroencephalogram-based negative emotion recognition method and system for aggressive behavior prediction | |
| Budak et al. | An effective hybrid model for EEG-based drowsiness detection | |
| Zhang et al. | Mindid: Person identification from brain waves through attention-based recurrent neural network | |
| Demir et al. | Exploring deep learning features for automatic classification of human emotion using EEG rhythms | |
| Radman et al. | Multi-feature fusion approach for epileptic seizure detection from EEG signals | |
| CN114052735B (en) | Deep field self-adaption-based electroencephalogram emotion recognition method and system | |
| Bhardwaj et al. | Classification of human emotions from EEG signals using SVM and LDA Classifiers | |
| Kalaivani et al. | Analysis of EEG signal for the detection of brain abnormalities | |
| Shahnaz et al. | Emotion recognition based on wavelet analysis of Empirical Mode Decomposed EEG signals responsive to music videos | |
| Nath et al. | An efficient approach to eeg-based emotion recognition using lstm network | |
| Abdollahpour et al. | Transfer learning convolutional neural network for sleep stage classification using two-stage data fusion framework | |
| An et al. | Electroencephalogram emotion recognition based on 3D feature fusion and convolutional autoencoder | |
| CN112200016A (en) | Electroencephalogram signal emotion recognition based on ensemble learning method AdaBoost | |
| CN109222966A (en) | A kind of EEG signals sensibility classification method based on variation self-encoding encoder | |
| CN115770044B (en) | Emotion recognition method and device based on electroencephalogram phase amplitude coupling network | |
| CN109871831A (en) | A kind of emotion identification method and system | |
| Wang et al. | Stimulus-stimulus transfer based on time-frequency-joint representation in SSVEP-based BCIs | |
| Lian et al. | The improved ELM algorithms optimized by bionic WOA for EEG classification of brain computer interface | |
| Xu et al. | EEG emotion classification based on baseline strategy | |
| CN115859185A (en) | Electroencephalogram emotion recognition method based on pulse convolution neural network | |
| Behnam et al. | Singular Lorenz Measures Method for seizure detection using KNN-Scatter Search optimization algorithm | |
| Tang et al. | Deep neural network for emotion recognition based on meta-transfer learning | |
| Kim et al. | eRAD-Fe: Emotion recognition-assisted deep learning framework | |
| Saha et al. | Automatic emotion recognition from multi-band EEG data based on a deep learning scheme with effective channel attention | |
| Dia et al. | A novel stochastic transformer-based approach for post-traumatic stress disorder detection using audio recording of clinical interviews |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| FGI | Letters patent sealed or granted (innovation patent) | ||
| MK22 | Patent ceased section 143a(d), or expired - non payment of renewal fee or expiry |