CN112215044A - Driving tendency identification method based on probabilistic neural network - Google Patents
Driving tendency identification method based on probabilistic neural network Download PDFInfo
- Publication number
- CN112215044A CN112215044A CN201910628973.2A CN201910628973A CN112215044A CN 112215044 A CN112215044 A CN 112215044A CN 201910628973 A CN201910628973 A CN 201910628973A CN 112215044 A CN112215044 A CN 112215044A
- Authority
- CN
- China
- Prior art keywords
- data
- driving
- neural network
- driving tendency
- layer
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/70—Multimodal biometrics, e.g. combining information from different biometric modalities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06N3/047—Probabilistic or stochastic networks
Abstract
The invention discloses a driving tendency identification method based on a probabilistic neural network, belongs to the field of intelligent transportation, comprehensively considers two aspects of psychology of a driver and driving data, and identifies the driving tendency of an automobile by using the probabilistic neural network. A driving tendency identification method based on a probabilistic neural network comprises the following steps: the method comprises the steps of carrying out static calibration on the driving tendency of a driver, collecting corresponding physiological data, collecting corresponding vehicle running data, training a probabilistic neural network and identifying the driving tendency. The driving tendency identification method based on the probabilistic neural network can identify the driving tendency of a driver according to the collected data. Through the identification of the driving tendency, the vehicle is subjected to more accurate early warning design, and the risk of accident occurrence is reduced.
Description
Technical Field
The invention relates to the field of intelligent traffic, in particular to a driving tendency identification method based on a probabilistic neural network.
Background
In recent years, with the rapid development of modern transportation industry, the rapid increase of automobile reserves brings great convenience to life and makes traffic safety issue increasingly prominent. How to reduce the occurrence of traffic accidents and ensure the safety of roads becomes a major hotspot of the current society. With the further development of research, the key points of automobile safety gradually turn from passive safety to active safety, wherein an automobile auxiliary driving system carries out early warning and other auxiliary measures on dangerous driving, and the automobile auxiliary driving system is greatly helpful for reducing traffic accidents.
Due to the difference of physiological-psychological characteristics of drivers, the current assistant driving system is not well suitable for each driver, and it is imperative to fully understand the driving behaviors of the drivers in order to improve the accuracy of the assistant driving system. The physiological-psychological characteristics of the driver of a vehicle can be expressed to a large extent by driving tendencies, which are expressed basically in 3 types at present: aggressive type, normal type, conservative type. The auxiliary driving system is designed according to the driving tendency type of the driver, so that the accuracy of the auxiliary driving system can be greatly improved. Therefore, research for accurately identifying driving tendency is necessary.
Disclosure of Invention
The invention provides a driving tendency identification method based on a probabilistic neural network.
A driving tendency identification method based on a probabilistic neural network comprises the following steps:
collecting driving data and physiological data corresponding to each driving tendency required by training a neural network: the driving tendency of a driver is statically calibrated in a meter self-testing mode, driving data corresponding to the driver are collected by using a simulation driving system, and physiological data corresponding to the driver are collected by using human factor equipment; using a PCA algorithm to perform dimensionality reduction on the acquired data, and calculating the most representative characteristic component; and training the probabilistic neural network by using the obtained characteristic data to obtain a neural network model capable of identifying the corresponding driving tendency according to the input data.
The invention relates to a driving tendency identification method based on a probabilistic neural network, which is characterized in that the related driving tendency is marked as follows:
the driving tendency of the driver is calibrated by measuring the self-measuring state of the scale, and the driving tendency is divided into an aggressive type (type 1), a normal type (type 2) and a conservative type (type 3) according to a threshold value specified by a questionnaire.
The invention relates to a driving tendency identification method based on a probabilistic neural network, which comprises the following steps:
the method comprises the following steps that a driver operates a simulation driver to collect driving data, wherein 9 characteristic parameters such as speed, acceleration, inter-vehicle distance, refueling frequency, braking frequency, refueling force, braking force, inter-vehicle insertable clearance and transverse spacing distance are counted;
the physiological data of the driver collected by human factor equipment, such as ECG (electrocardiogram), EDA (skin electric), EMG (electromyogram), PPG (photoplethysmography), RESP (respiration) and SKT (skin temperature), are used for totaling 6 characteristic parameters.
The invention relates to a driving tendency identification method based on a probabilistic neural network, which relates to a method for reducing the dimension of measured data, and comprises the following steps:
using PCA algorithm to extract main features for dimensionality reduction, and the specific flow is as follows:
1) standardizing the original 15-dimensional data set;
2) a covariance matrix of the samples is constructed. Suppose thatX=(x 1 , x 2 ,…,x 288 )TIs composed of288*15Wherein 288 is the number of data sets, 15 is the number of characteristic parameters, and the covariance calculation formula is as follows:
wherein the content of the first and second substances,
is as followsj,kAn expected value of each characteristic parameter;
3) calculating an eigenvalue of the covariance matrix and a corresponding eigenvector;
4) the eigenvector corresponding to the first 5 largest eigenvalues is selected. The corresponding eigenvalue magnitude determines the importance (i.e., contribution ratio) of the eigenvector, and the contribution ratio formula is:
the contribution rates of the first 5 maximum eigenvalues are accumulated to 85%, namely the first 5 eigenvectors can represent the original data;
5) constructing a mapping matrix W through the first 5 eigenvectors;
6) the 15-dimensional input data set X is transformed into a new 5-dimensional feature subspace by means of a mapping matrix W, resulting in 5 principal components.
The invention relates to a driving tendency identification method based on a probabilistic neural network, wherein the probabilistic neural network (the structure is shown as a figure 2) is as follows:
1) the input layer is provided with 6 nodes which represent the number of input characteristic parameters and are used for receiving values from training samples and transmitting data to the hidden layer;
2) the hidden layer is a radial base layer and has 288 nodes, representing the 288 samples selected. The input-output relational formula of the jth node of the ith type mode is as follows:
where i =1,2,3, i.e. three driving tendencies.
Is the jth center of class i;
3) the summation layer has 3 nodes which represent 3 types of driving tendencies, the layer carries out weighted summation on the outputs of the nodes belonging to the same class in the hidden layer, and the formula is as follows:
v i l represents the number of nodes of the ith class;
4) the output layer is composed of 3 competitive neurons, receives the output of the summation layer, finds a node corresponding to the value with the maximum posterior probability density in the output layer, and the formula is as follows:
the node output is 1, namely the measured data is judged to belong to the driving tendency type corresponding to the node, and the node outputs of other types are 0.
Drawings
FIG. 1 is a flow chart of driving tendency recognition model training.
Fig. 2 is a diagram of a probabilistic neural network architecture.
Detailed Description
The present invention will be described in detail with reference to the following embodiments.
There are individual differences from person to person, but according to the study of cognitive psychology, certain characteristics of a person can be represented to a large extent in several patterns. Each driver has different physiological-psychological characteristics, which can be roughly expressed by three driving tendencies: aggressive type, normal type, conservative type.
The invention comprehensively considers the external expression of each driving tendency in the driving process, namely driving data and physiological data (such as electrocardio), and the two aspects represent the driving characteristics of the driver together and are used as data classification bases.
And (3) performing static calibration on the driving tendency in a gauge self-testing mode, and judging classification according to a threshold value.
The initial characteristic data of each driving tendency collected by the invention comprises 15 driving characteristic parameters of speed, acceleration, inter-vehicle distance, refueling frequency, braking frequency, refueling force, braking force, inter-vehicle insertable space and transverse spacing distance, and 6 psychological characteristic parameters of ECG (electrocardiogram), EDA (skin electricity), EMG (electromyography), PPG (photoplethysmography), RESP (respiration) and SKT (skin temperature).
In practical application, data acquisition equipment is used for acquiring relevant data in driving in real time, and the acquired data is imported into a model for dimension reduction.
The invention uses PCA algorithm to extract main characteristics for dimension reduction,
1) the original 15-dimensional dataset is first normalized. Then constructing a covariance matrix of the sample;
suppose thatX=(x 1 , x 2 ,…,x 288 )TIs composed of288*15Wherein 288 is the number of data sets, 15 is the number of characteristic parameters, and the covariance calculation formula is as follows:
wherein the content of the first and second substances,
is as followsj,kAn expected value of each characteristic parameter;
2) calculating an eigenvalue of the covariance matrix and a corresponding eigenvector; the eigenvector corresponding to the first 5 largest eigenvalues is selected. The corresponding eigenvalue magnitude determines the importance (i.e., contribution ratio) of the eigenvector, and the contribution ratio formula is:
the contribution rates of the first 5 maximum feature values are accumulated to 85 percent, namely, the first 5 feature vectors can represent the original data;
3) constructing a mapping matrix W through the first 5 eigenvectors;
4) the 15-dimensional input data set X is transformed into a new 5-dimensional feature subspace by means of a mapping matrix W, resulting in 5 principal components.
And inputting the data subjected to the dimensionality reduction into a probabilistic neural network for training to obtain a model capable of identifying the driving tendency.
1. Theoretical basis of probabilistic neural network:
the probabilistic neural network is established on the basis of Bayes minimum risk criterion, firstly, L = A or B is assumed, and prior probabilities are respectively HA、HBAnd H isA+HB= 1; given an input vector ofx=[x1,x2,…, xN]，
In thatxIn the event of occurrence, the posterior probability of class L is
The classification is based on:
when classifying, the input vector should be classified into the class with the higher posterior probability.
2. The probabilistic neural network identification structure:
1) the input layer is provided with 5 nodes which represent the number of input characteristic parameters and are used for receiving values from training samples and transmitting data to the hidden layer;
2) the hidden layer is a radial base layer and has 288 nodes, representing the 288 samples selected. The input-output relational formula of the jth node of the ith type mode is as follows:
where i =1,2,3, i.e. three driving tendencies.
Is the jth center of class i;
3) the summation layer has 3 nodes which represent 3 types of driving tendencies, the layer carries out weighted summation on the outputs of the nodes belonging to the same class in the hidden layer, and the formula is as follows:
4) the output layer is composed of 3 competitive neurons, receives the output of the summation layer, finds a node corresponding to the value with the maximum posterior probability density in the output layer, and the formula is as follows:
the node output is 1, namely the measured data is judged to belong to the driving tendency type corresponding to the node, and the node outputs of other types are 0.
Claims (4)
1. A driving tendency identification method based on a probabilistic neural network is established according to the following steps:
1) collecting driving data and physiological data corresponding to each driving tendency required by training a neural network: the driving tendency is statically calibrated in a meter self-testing mode, the driving data corresponding to a driver are collected by using a simulation driving system, and the physiological data corresponding to the driver are collected by using human factor equipment;
2) using a PCA algorithm (principal component analysis) to reduce the dimension of the acquired data, and screening out the most representative characteristic value;
3) and training the obtained characteristic data based on the probabilistic neural network to obtain a driving tendency identification model.
2. The data collection of claim 1, wherein:
1) the static calibration of the driving tendency is carried out by adopting a meter self-testing form and is divided into an aggressive type, a common type and a conservative type;
2) the driving data is collected by a driver operating simulation driver, and the measured data comprises: the speed, the acceleration, the vehicle distance, the refueling frequency, the braking frequency, the refueling force, the braking force, the vehicle-insertable gap between the vehicles and the transverse spacing distance are 9 characteristic parameters in total;
3) the physiological data is collected by using human factor equipment, and the measured data comprises the following data: the total of 6 characteristic parameters are ECG (electrocardiogram), EDA (skin electric), EMG (electromyogram), PPG (photoplethysmography), RESP (respiration) and SKT (skin temperature).
3. The dimension reduction process according to claim 1, characterized by: using PCA algorithm to extract main features for dimensionality reduction, and the specific flow is as follows:
1) standardizing the original 15-dimensional data;
2) a covariance matrix of the samples is constructed,
3) calculating an eigenvalue of the covariance matrix and a corresponding eigenvector;
4) selecting the eigenvectors corresponding to the first a maximum eigenvalues, wherein the importance (contribution rate) of the eigenvectors is determined by the size of the corresponding eigenvalue, and the cumulative sum of the contribution rates of the first 5 maximum eigenvalues reaches 85%, namely the first 5 eigenvectors can represent the original data;
5) constructing a mapping matrix W through the first 5 eigenvectors;
6) the 15-dimensional input data set X is transformed into a new 5-dimensional feature subspace by means of a mapping matrix W, resulting in 5 principal components.
4. The probabilistic neural network model of claim 1, wherein: the probabilistic neural network model consists of an input layer, a hidden layer, a summation row and an output layer, and the specific model is as follows:
1) the input layer is provided with 5 nodes which represent the number of input characteristic parameters and are used for receiving values from training samples and transmitting data to the hidden layer;
2) the hidden layer is a radial base layer, and has 288 nodes which represent the 288 samples selected;
3) the summation layer has 3 nodes which represent 3 types of driving tendencies, and the summation layer carries out weighted summation on the outputs of the nodes belonging to the same type in the hidden layer;
4) the output layer is composed of 3 competitive neurons, receives the output of the summation layer, finds a node corresponding to the value with the maximum posterior probability density in the node, the output of the node is 1, namely the measured data is judged to belong to the driving tendency type corresponding to the node, and the output of the nodes of other types is 0.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910628973.2A CN112215044A (en) | 2019-07-12 | 2019-07-12 | Driving tendency identification method based on probabilistic neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910628973.2A CN112215044A (en) | 2019-07-12 | 2019-07-12 | Driving tendency identification method based on probabilistic neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112215044A true CN112215044A (en) | 2021-01-12 |
Family
ID=74047760
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910628973.2A Pending CN112215044A (en) | 2019-07-12 | 2019-07-12 | Driving tendency identification method based on probabilistic neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112215044A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113392892A (en) * | 2021-06-08 | 2021-09-14 | 重庆大学 | Method and device for identifying driving habits of driver based on data fusion |
| CN113408431A (en) * | 2021-06-22 | 2021-09-17 | 青岛小鸟看看科技有限公司 | Intelligent driving evaluation training method and system based on eyeball tracking |
| CN115034337A (en) * | 2022-08-10 | 2022-09-09 | 江西科骏实业有限公司 | Method and device for identifying state of traction motor in rail transit vehicle and medium |
-
2019
- 2019-07-12 CN CN201910628973.2A patent/CN112215044A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113392892A (en) * | 2021-06-08 | 2021-09-14 | 重庆大学 | Method and device for identifying driving habits of driver based on data fusion |
| CN113408431A (en) * | 2021-06-22 | 2021-09-17 | 青岛小鸟看看科技有限公司 | Intelligent driving evaluation training method and system based on eyeball tracking |
| CN115034337A (en) * | 2022-08-10 | 2022-09-09 | 江西科骏实业有限公司 | Method and device for identifying state of traction motor in rail transit vehicle and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113627518B (en) | Method for realizing neural network brain electricity emotion recognition model by utilizing transfer learning | |
| CN109124625B (en) | Driver fatigue state level grading method | |
| CN112215044A (en) | Driving tendency identification method based on probabilistic neural network | |
| CN104809226B (en) | A kind of method of the uneven multivariate time series data of early stage classification | |
| US20220319706A1 (en) | A drgs automatic grouping method based on a convolutional neural network | |
| Suzdaleva et al. | An online estimation of driving style using data-dependent pointer model | |
| CN108416364A (en) | Integrated study data classification method is merged in subpackage | |
| CN111242484A (en) | Vehicle risk comprehensive evaluation method based on transition probability | |
| Li et al. | Medical data stream distribution pattern association rule mining algorithm based on density estimation | |
| CN108769104B (en) | Road condition analysis and early warning method based on vehicle-mounted diagnosis system data | |
| CN113484738A (en) | Circuit fault diagnosis method based on multi-feature information fusion | |
| CN115828140A (en) | Neighborhood mutual information and random forest fusion fault detection method, system and application | |
| CN111081379A (en) | Disease probability decision method and system | |
| CN110929224A (en) | Safety index system establishing method based on bus driving safety | |
| CN111583592B (en) | Experimental environment safety early warning method based on multidimensional convolution neural network | |
| CN105303771A (en) | Fatigue judging system and method | |
| Zhu et al. | A driving behavior risk classification framework via the unbalanced time series samples | |
| CN111931663A (en) | Fluorescence immunochromatography quantitative image peak-finding algorithm based on deep learning | |
| CN111227826B (en) | Electroencephalogram signal analysis method based on network motif | |
| CN108509999A (en) | It is a kind of indignation drive detection and safe early warning method | |
| CN111709441A (en) | Behavior recognition feature selection method based on improved feature subset discrimination | |
| CN114464319B (en) | AMS susceptibility assessment system based on slow feature analysis and deep neural network | |
| CN111209816B (en) | Non-contact fatigue driving detection method based on regular extreme learning machine | |
| CN112690774B (en) | Magnetic resonance image-based stroke recurrence prediction method and system | |
| CN110704789B (en) | Population dynamic measurement and calculation method and system based on 'urban superconcephalon' computing platform |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20210112 |