CN109447182A - Driving behavior classification method and device based on HMM algorithm - Google Patents

Driving behavior classification method and device based on HMM algorithm Download PDF

Info

Publication number
CN109447182A
CN109447182A CN201811378191.XA CN201811378191A CN109447182A CN 109447182 A CN109447182 A CN 109447182A CN 201811378191 A CN201811378191 A CN 201811378191A CN 109447182 A CN109447182 A CN 109447182A
Authority
CN
China
Prior art keywords
driving
model
dangerous
likelihood value
normal driving
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
Application number
CN201811378191.XA
Other languages
Chinese (zh)
Inventor
刘均
于海悦
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen Launch Technology Co Ltd
Original Assignee
Shenzhen Launch Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shenzhen Launch Technology Co Ltd filed Critical Shenzhen Launch Technology Co Ltd
Priority to CN201811378191.XA priority Critical patent/CN109447182A/en
Publication of CN109447182A publication Critical patent/CN109447182A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G06F18/24155—Bayesian classification
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Traffic Control Systems (AREA)

Abstract

This application discloses a kind of driving behavior classification methods based on HMM algorithm, comprising: obtains and drives observation data；The normal driving likelihood value for driving observation data under normal driving model is calculated to forward algorithm using preceding, and dangerous driving likelihood value of the driving behavior data under the dangerous driving model is calculated using the forward algorithm；Bayesian Factor is determined according to the normal driving likelihood value and the dangerous driving likelihood value；Determining driving behavior classification is compared with threshold value according to the Bayesian Factor, the accuracy of identification driving behavior can be improved by implementing the application.

Description

Driving behavior classification method and device based on HMM algorithm
Technical field
This application involves computer field more particularly to a kind of driving behavior classification methods and device based on HMM algorithm.
Background technique
Lane-change is most common in driving procedure while being also the higher driving behavior of degree of danger.According to United States highways Safety management bureau data shows, the traffic accident caused due to lane-change process accounting in the traffic accident of all statistics is high Up to 27%.On the basis of to Ben Che and the perception of surrounding vehicles operating status, the prediction technique of dangerous lane-change driving behavior is studied, Help to realize DAS (Driver Assistant System) accurately and timely lane-change early warning or intervention.Existing a large amount of lane-change study of warning be with Collision time based on speed and relative distance, or based on the minimum safe spacing of vehicle braking kinematics analysis as early warning ginseng Number, by determining that the threshold value of early-warning parameters establishes different prediction policies.And actually, dangerous lane-change is since lane-change to generation The whole process of danger conflict is difficult to be described with single early-warning parameters, needs to carry out using more complicated algorithm and model Research.For this reason, having chosen supporting vector machine model algorithm herein, the dangerous lane-change driving behavior prediction based on algorithm is established Model.At present, SVM has been achieved in terms of identification (is such as kept straight on, turned to, changing Lane) in driver intention/behavior good Prediction effect is simultaneously widely used in vehicle DAS (Driver Assistant System), but in terms of lane-change driving behavior danger forecasting research compared with It is few.
Summary of the invention
The embodiment of the present application technical problem to be solved is, provides the classification method and device of a kind of driving behavior, Can be effectively predicted whether lane-change behavior is hazardous act.
In order to solve the above-mentioned technical problem, this application provides a kind of driving behavior classification method based on HMM algorithm, packets It includes: obtaining and drive observation data, driving observation data is the collected sequence of observations in certain period of time, in the sequence of observations Each element include acquisition time.It is normal under normal driving model that driving observation data are calculated using forward algorithm Likelihood value is driven, and calculates the dangerous driving for driving observation data under dangerous driving model using the forward algorithm Likelihood value determines Bayesian Factor according to the normal driving likelihood value and the dangerous driving likelihood value, according to Bayes because Son and threshold value, which are compared, to be determined to drive and observes the corresponding driving behavior classification of data.
The embodiment of the present application respectively obtains seemingly under normal driving model and dangerous driving model to observation data are driven So value is compared with threshold value by the Bayesian Factor obtained to two likelihood values and differentiates driving behavior classification, can improve knowledge The accuracy of other driving behavior, improves the safety of driving.
It in a kind of possible design, obtains before driving observation data, further includes: training set is run according to normal driving Normal driving model is established using HMM algorithm, and training set is run according to dangerous driving, dangerous driving is established using HMM algorithm Model.Wherein, normal driving operation training set includes multiple normal driving behavior sample data, and dangerous driving runs training set packet Multiple dangerous driving behavior sample datas are included, the corresponding driving behavior classification of dangerous driving behavior sample data is dangerous driving row For the corresponding driving behavior classification of normal driving behavior sample data is normal driving behavior, dangerous driving behavior sample data It is a time series with normal driving behavior sample data.
In a kind of possible design, in the normal driving operation training set and dangerous driving operation training set Data include that the lateral velocity between adjacent two vehicle is poor, the longitudinal velocity between adjacent two vehicle is poor, the transverse direction between adjacent two vehicle Acceleration is poor, the longitudinal acceleration between adjacent two vehicle is poor, the traffic direction angle between adjacent two vehicle, between adjacent two vehicle One of centroid distance is a variety of.
In a kind of possible design, normal driving model and dangerous driving model and hidden state sequence, sample are observed Value sequence, state transition probability, the probability distribution function of state output event are related with initial state distribution.
The model posterior probability for driving observation dataI=0 or 1, z are described drive Sail observation data, and P (z | λ0) the expression normal driving likelihood value, p (z) the expression probability for driving observation data, P (z | λ1) indicate the dangerous driving likelihood value, the Bayesian Factor
Second aspect, this application provides a kind of sorters of driving behavior, comprising:
Acquiring unit drives observation data for obtaining；
Computing unit observes data under normal driving model just for calculating drivings using forward algorithm Likelihood value is often driven, and the driving behavior data are calculated under the dangerous driving model using the forward algorithm Dangerous driving likelihood value；
The computing unit is also used to determine pattra leaves according to the normal driving likelihood value and the dangerous driving likelihood value This factor；
Taxon, for being compared determining driving behavior classification with threshold value according to the Bayesian Factor.
In a kind of possible design, further includes:
Training unit establishes normal driving model, Yi Jigen using HMM algorithm for running training set according to normal driving Dangerous driving model is established using HMM algorithm according to dangerous driving operation training set.
In a kind of possible design, in the normal driving operation training set and dangerous driving operation training set Data include that the lateral velocity between adjacent two vehicle is poor, the longitudinal velocity between adjacent two vehicle is poor, the transverse direction between adjacent two vehicle Acceleration is poor, the longitudinal acceleration between adjacent two vehicle is poor, the traffic direction angle between adjacent two vehicle, between adjacent two vehicle One of centroid distance is a variety of.
In a kind of possible design, the normal driving model and the dangerous driving model and hidden state sequence, Sample observations sequence, state transition probability, the probability distribution function of state output event are related with initial state distribution.
In a kind of possible design, the model posterior probability for driving observation dataI=0 or 1, z be drivings observation data, P (z | λ0) indicate the normal driving likelihood Value, p (z) the expression probability for driving observation data, P (z | λ1) indicate the dangerous driving likelihood value, the Bayes because Son
The another aspect of the application provides a kind of device, and the classification side of the driving behavior of above-mentioned first aspect may be implemented Method.Such as described device can be chip (such as baseband chip or communication chip etc.) or processing server.It can be by soft Part, hardware or the corresponding software realization above method is executed by hardware.
It in one possible implementation, include processor, memory in the structure of described device；The processor quilt It is configured to that described device is supported to execute corresponding function in above-mentioned communication means.Memory is saved for coupling with processor The necessary program of described device (instruction) and/or data.Optionally, the communication device can also include communication interface for branch Hold the communication between described device and other network elements.
In alternatively possible implementation, described device may include the list for executing corresponding actions in the above method Element module.
In another possible implementation, including processor and R-T unit, the processor and the transmitting-receiving fill Set coupling, the processor for executing computer program or instruction, with control the R-T unit carry out information reception and It sends；When the processor executes the computer program or instruction, the processor is also used to realize the above method.Its In, the R-T unit can be transceiver, transmission circuit or input/output interface.When the communication device is chip, institute Stating R-T unit is transmission circuit or input/output interface.
When described device is chip, transmission unit can be output unit, such as output circuit or communication interface；It connects Receiving unit can be input unit, such as input circuit or communication interface.When the communication device is the network equipment, send Unit can be transmitter or transmitter；Receiving unit can be receiver or receiver.
The another aspect of the application provides a kind of device, which includes: memory and processor；Wherein, the storage Batch processing code is stored in device, and the processor executes each side for calling the program code stored in the memory Method described in face.
The another aspect of the application has been mentioned for a kind of computer readable storage medium, in the computer readable storage medium It is stored with instruction, when run on a computer, so that computer executes method described in above-mentioned various aspects.
The another aspect of the application provides a kind of computer program product comprising instruction, when it runs on computers When, so that computer executes method described in above-mentioned various aspects.
Detailed description of the invention
In order to illustrate the technical solutions in the embodiments of the present application or in the prior art more clearly, to embodiment or will show below There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this Some embodiments of application for those of ordinary skill in the art without creative efforts, can be with It obtains other drawings based on these drawings.
Fig. 1 is a kind of flow diagram of driving behavior classification method based on HMM algorithm provided by the embodiments of the present application；
Fig. 2 a is a kind of another process of the classification method of driving behavior based on HMM algorithm provided by the embodiments of the present application Schematic diagram；
Fig. 2 b is the distribution schematic diagram of forecast interval provided by the embodiments of the present application；
Fig. 2 c is the functional block diagram of the driving behavior classification method provided by the embodiments of the present application based on HMM algorithm；
Fig. 3 is a kind of another structural schematic diagram of device provided by the embodiments of the present application；
Fig. 4 is a kind of another structural schematic diagram of device provided by the embodiments of the present application.
Specific embodiment
It is with reference to the accompanying drawings and embodiments, right in order to which the objects, technical solutions and advantages of the application are more clearly understood The application is further elaborated.It should be appreciated that specific embodiment described herein is only used to explain the application, and It is not used in restriction the application.Meanwhile in the description of the present application, term " first ", " second " etc. are only used for distinguishing description, without It can be interpreted as indication or suggestion relative importance.It will be clear to one skilled in the art that these details its The application also may be implemented in its embodiment.In other situations, it omits to well-known system, device, circuit and side The detailed description of method, so as not to obscure the description of the present application with unnecessary details.
Understood for the ease of the embodiment of the present application, below to the invention relates to relational language explain Illustrate:
Maximum entropy model (EM), entropy have measured the uncertainty of things, more uncertain things, and entropy is bigger.At random The expression formula of the entropy of variable X is as follows:
Wherein, n indicates the different discrete value of the n kind of X, piExpression X value is i Probability, log indicates using 2 to be bottom or e as the logarithm at bottom.
The combination entropy expression formula of two variable Xs and Y:
Conditional entropy is similar to conditional probability, and measurement Y remaining uncertainty, expression formula in the case where known X is as follows:
It is a conditional probability distribution P (Y | X) that maximum entropy model, which drives disaggregated model, and X is characterized, and Y is output.Given one A training set (x1,y1),(x2,y2),...,(xm,ym), wherein x is n dimensional feature vector, and y is classification output.Our target is A best classification type is selected with maximum entropy model.In the case where given training set, available totality Joint Distribution P The distribution of the experience of (X, Y) and edge distribution P (X) As in training set X and Y simultaneously occur number divided by Total sample number m,For in training set X occur appearance number divided by total sample number m.It is described with characteristic function f (x, y) It inputs x and exports the relationship between y, x and when meeting specified relationship with y is defined as f (x, y)=1；Otherwise f (x, y)=0, can As long as to think (the x occurred in training seti,yi), f (xi,yiThe same training sample in)=1 can have multiple constraints special Levy function.Characteristic function f (x, y) is distributed about experienceDesired valueFeature letter Number f (x, y) is distributed about condition distribution P (Y | X) and experienceDesired value
If can be from training focusing study to model, it is assumed that two desired values are equal, i.e.,
Conditions described above is the constraint condition of maximum entropy model study, if we have M characteristic function fi(x, Y), i=1,2,3 ..., M have M constraint condition.If can be understood as training set has M sample, just have and M sample Corresponding M constraint condition obtains maximum entropy model in this way and is defined as follows: assuming that meeting the model set of institute's Prescribed Properties ForI=1,2,3 ..., M, the conditional entropy being defined on conditional probability distribution P (Y | X) are as follows:Corresponding P when in order to keep H (P) maximum (y | x), can be to H (P) plus negative Number minimizing, in order to which making-H (P) is convex function, and the method for convex optimization is easy to use to seek extreme value.
Hidden Markov model (Hidden Markov Model, HMM) is statistical model, it is used to describe one to contain The Markov process of implicit unknown parameter.Its difficult point is the implicit parameter that the process is determined from the parameter of observable.Then Use these parameters to for further analysis, such as pattern-recognition.Hidden Markov model is one kind of Markov chain, it State cannot observe directly, but can be arrived by observation vector sequence inspection, each observation vector is by certain probability Density Distribution shows as various states, each observation vector is produced by a status switch with corresponding probability density distribution It is raw.So hidden Markov model is a dual random process, i.e., with the Hidden Markov Chain and display of certain status number Random function collection.
In order to illustrate technical solution described herein, the following is a description of specific embodiments.
It is a kind of stream of driving behavior classification method based on HMM algorithm provided by the embodiments of the present application referring to Fig. 1, Fig. 1 Journey schematic diagram, in the embodiment of the present application, which comprises
S101, it obtains to drive and observes data.
Specifically, driving observation device for classifying data collected time series in forecast interval, forecast interval is risen At the time of moment beginning is that occurs for vehicle more than preset duration the time of lateral displacement, the finish time of forecast interval is more than for vehicle At the time of adjacent lane line.Such as: according to predetermined period, the collected data in forecast interval are used as driving to sorter Observe data.
S102, the normal driving likelihood value for driving observation data under normal driving model is calculated using forward algorithm, And the dangerous driving likelihood value for driving observation data under dangerous driving model is calculated using forward algorithm.
Specifically, normal driving model is a kind of model of probability density distribution, indicate that driving observation data is normally driving Sail the probability of behavior.Dangerous driving model is also a kind of model of probability density distribution, indicates that it is dangerous for driving observation data The probability of driving behavior.Normal driving model and dangerous driving model are that sorter trains to come previously according to sample data 's.Sorter calculates the normal driving likelihood value for driving observation data under normal driving model using forward algorithm, with And drive dangerous driving likelihood value of the observation data under dangerous driving model.
S103, Bayesian Factor is determined according to normal driving likelihood value and dangerous driving likelihood value.
Specifically, sorter calculates the Bayesian Factor of above-mentioned two model using Bayesian model selection method, i.e., Bayesian Factor is determined according to normal driving likelihood value and dangerous driving likelihood value.
S104, determining driving behavior classification is compared according to Bayesian Factor and threshold value.
Specifically, preset threshold value is compared by sorter with Bayesian Factor, it is greater than threshold value in Bayesian Factor When, determine that driving the corresponding driving behavior classification of observation data is dangerous driving behavior, is less than or equal to threshold in Bayesian Factor When value, determine that driving the corresponding driving behavior classification of observation data is normal driving behavior.
Implement embodiments herein, driving observation data are obtained respectively under normal driving model and dangerous driving model The likelihood value arrived is compared with threshold value by the Bayesian Factor obtained to two likelihood values and differentiates driving behavior classification, energy The accuracy for improving identification driving behavior, improves the safety of driving.
It referring to fig. 2, is a kind of another stream of the driving behavior classification method based on HMM algorithm provided by the embodiments of the present application Journey schematic diagram, in the embodiment of the present application, which comprises
S201, normal driving model is established using HMM algorithm according to normal driving operation training set, and is driven according to danger It sails operation training set and dangerous driving model is established using HMM algorithm.
Specifically, normal driving operation training set includes multiple normal driving behavior sample data, normal driving behavior sample The corresponding driving behavior classification of notebook data is normal driving behavior, and normal driving behavior sample data are acquired in forecast interval Time series, at the time of the initial time of forecast interval is that vehicle occurs time of lateral displacement and is more than preset duration, prediction At the time of the finish time in section is that vehicle is more than adjacent lane line.Such as: driving behavior sample data be (x1, t1), (x2, t2) ..., (xn, tn), (xn, tn) indicate in tn moment collected data xn.It should be noted that different is normal The range of the corresponding forecast interval of driving behavior sample data may not be identical.
It includes multiple dangerous driving behavior sample datas, dangerous driving behavior sample data pair that dangerous driving, which runs training set, The driving behavior classification answered is dangerous driving behavior, and dangerous driving behavior sample data is the time sequence acquired in pre-set interval Column, the definition of pre-set interval is referring to above-mentioned explanation, the model of the different corresponding forecast intervals of dangerous driving behavior sample data Enclosing may not be identical.
Wherein, the process for establishing dangerous driving model and normal driving model is illustrated below:
The pdf model for establishing sample makes inferences prediction using model.Probability base of the HMM algorithm to observation time Assume in following: observable event determines to meet general random process by a series of limited unobservable hidden states, Mutual conversion process between hidden state meets Markov random process, i.e. dual random process.In view of driving for vehicle The data that observation data are continuous type are sailed, and gauss hybrid models (GMM) can infinitely approach the distribution of arbitrary continuation type variable, Therefore the probability distribution function of state output event is established using GMM.HMM model (dangerous driving model based on GMM description output Or normal driving model) can state are as follows: λ={ Π, A, c, μ, U }.Wherein, crucial concept and meaning of parameters are:
(1) hidden state sequence: unobservable implicit sequence Q=corresponding with one sample data (time series) q1,q2,q3,…,qT, wherein each hidden state in implicit sequence is from finite aggregate S, S=s comprising N number of state1, s2,s3,…,sN。
(2) sample observations sequence: the sample data O, O=o directly observed in forecast interval1,o2, o3,…,oT.Each sample value ot∈Rd, indicate the vehicle operation characteristic vector of the collected d dimension of t moment, in the present embodiment, Vehicle operation characteristic vector includes that the longitudinal velocity between two vehicles that the lateral velocity between two adjacent vehicles is poor, adjacent is poor, phase Transverse acceleration between two adjacent vehicles is poor, the direction between two poor, adjacent vehicles of the longitudinal acceleration between two adjacent vehicles Centroid distance between angle, two adjacent vehicles.
(3) state transition probability: αij, i and j={ 1,2,3 ..., N }, αijExpression system is by state SiTo another state Sj Transition probability, the transition probability between different conditions constitutes shift-matrix A.
(4) probability distribution function of state output event:Wherein N (μjm,Ujm) be output valve under state j multidimensional Gaussian density function, μjmFor mean vector, UjmFor covariance matrix, M is Gauss Mix number, cjmFor Gaussian Mixture coefficient,
(5) initial state distribution: ΠT=[π1,π2,π3,...,πN], wherein πiIndicate that initial hidden state is SiIt is general Rate.
Data are observed in S202, the driving obtained in forecast interval.
Specifically, the definition of forecast interval is referring to the definition in S201, details are not described herein again, and sorter can be according to default Period acquire multiple data in forecast interval and formed and drive observation data.Wherein, the data type of acquisition can be adjacent Two vehicles between poor, adjacent two vehicles of lateral velocity between poor, adjacent two vehicles of longitudinal velocity between transverse acceleration Difference, the mass center between angular separation, two adjacent vehicles between two poor, adjacent vehicles of longitudinal acceleration between two adjacent vehicles Distance.
S203, the normal driving likelihood value for driving observation data under normal driving model is calculated using forward algorithm, And dangerous driving likelihood value of the measuring behavior data under dangerous driving model is calculated using forward algorithm.
Specifically, normal driving model is a kind of model of probability density distribution, indicate that driving observation data is normally driving Sail the probability of behavior.Dangerous driving model is also a kind of model of probability density distribution, indicates that it is dangerous for driving observation data The probability of driving behavior.Normal driving model and dangerous driving model are that sorter trains to come previously according to sample data 's.Sorter calculates the normal driving likelihood value for driving observation data under normal driving model using forward algorithm, with And drive dangerous driving likelihood value of the observation data under dangerous driving model.Such as: obtain normal driving model λ0And danger Driving model λ1Afterwards, it gives arbitrary drive and observes data z, then can calculate separately normal driving likelihood by forward algorithm Value P (z | λ0) and dangerous driving likelihood value P (z | λ1)。
Wherein, before calculating normal driving likelihood value and dangerous driving likelihood value, the step of data initialization is needed to be implemented Suddenly, the process of data initialization is described below:
Normal driving model and danger are driven using expectation maximization (expectation-maximization, EM) algorithm Model is sailed to be learnt to obtain the estimation of maximum likelihood value.For A and Π, the method that random or uniform value can be used carries out initial Change.C, U and μ can be gathered sample data according to status number N and Gaussian mixture number M using K-means clustering algorithm automatically For N × M class, to obtain the initial value of c, U and μ.
Such as: class representated by m-th of Gauss member inside note state n is lnm, according to sample data { oi, i=1, 2,...,nsampleCalculate the initial value of c, U and μ.
Wherein, as x ∈ 1, δ (x, l)=1.WhenWhen, δ (x, l)=0.X is to drive observation data, and l is cluster Serial number.
S204, Bayesian Factor is determined according to normal driving likelihood value and dangerous driving likelihood value.
Specifically, comparing two models using the calculating of Bayesian model selection method to improve the precision of prediction of model Bayesian Factor, and driving behavior classification is obtained after carrying out threshold value comparison to it.According to Bayes' theorem, it is known that drive observation number According to model after prolong probabilityThe pattra leaves of dangerous driving model and normal driving model This factor B F meets:Drive the prior probability P (λ of observation data1)=P (λ0)=0.5, can be according to elder generation Test probability and posterior probability setting threshold value.Log-likehood can also be used to calculate Bayesian Factor.
S205, judge whether Bayesian Factor is greater than threshold value.
S206, determine that driving the corresponding driving behavior classification of observation data is dangerous driving behavior.
Specifically, determining that the driving behavior classification for driving observation data is driven for danger when Bayesian Factor is greater than threshold value Behavior is sailed,
S207, determine that driving the corresponding driving behavior classification of observation data is normal driving behavior.
Specifically, determining that driving the corresponding driving behavior classification of observation data is when Bayesian Factor is not more than threshold value Normal driving behavior.
Implement embodiments herein, driving observation data are obtained respectively under normal driving model and dangerous driving model The likelihood value arrived is compared with threshold value by the Bayesian Factor obtained to two likelihood values and differentiates driving behavior classification, energy The accuracy for improving identification driving behavior, improves the safety of driving.
Above-mentioned Fig. 2 illustrates a kind of driving behavior classification method based on SVM algorithm of the embodiment of the present application.
Fig. 3 is referred to, Fig. 3 is a kind of structural schematic diagram of device provided by the embodiments of the present application, which may include Acquiring unit 301, computing unit 302 and taxon 303.
Acquiring unit 301 drives observation data for obtaining.
Computing unit 302, for calculating the driving observation data under normal driving model using forward algorithm Normal driving likelihood value, and the driving behavior data are calculated in the dangerous driving model using the forward algorithm Under dangerous driving likelihood value.
Computing unit 302 is also used to determine pattra leaves according to the normal driving likelihood value and the dangerous driving likelihood value This factor.
Taxon 303, for being compared determining driving behavior classification with threshold value according to the Bayesian Factor.
In a kind of possible embodiment, device 3 further include:
Training unit establishes normal driving model, Yi Jigen using HMM algorithm for running training set according to normal driving Dangerous driving model is established using HMM algorithm according to dangerous driving operation training set.
In a kind of possible embodiment, the normal driving operation training set and the dangerous driving run training set In data include that lateral velocity between adjacent two vehicle is poor, the longitudinal velocity between adjacent two vehicle is poor, between adjacent two vehicle Transverse acceleration is poor, the longitudinal acceleration between adjacent two vehicle is poor, the traffic direction angle between adjacent two vehicle, adjacent two vehicle it Between one of centroid distance or a variety of.
In a kind of possible embodiment, the normal driving model and the dangerous driving model and hidden state sequence Column, the probability distribution function of sample observations sequence, state transition probability, state output event are related with initial state distribution.
In a kind of possible embodiment, the model posterior probability for driving observation dataI=0 or 1, z be drivings observation data, P (z | λ0) indicate the normal driving likelihood Value, p (z) the expression probability for driving observation data, P (z | λ1) indicate the dangerous driving likelihood value, the Bayes because Son
Device 3 can be terminal device, described device 3 or the field programmable gate array for realizing correlation function (field-programmable gate array, FPGA), special integrated chip, System on Chip/SoC (system on chip, SoC), central processing unit (central processor unit, CPU), network processing unit (network processor, NP), Digital signal processing circuit, microcontroller (micro controller unit, MCU), can also use programmable controller (programmable logic device, PLD) or other integrated chips.
The embodiment of the present application and the embodiment of the method for Fig. 2 a are based on same design, and bring technical effect is also identical, tool Body process can refer to the description of the embodiment of the method for Fig. 2 a, and details are not described herein again.
Fig. 4 is a kind of apparatus structure schematic diagram provided by the embodiments of the present application, and hereinafter referred to as device 4, device 4 can integrate In terminal device above-mentioned, as shown in figure 4, the device includes: memory 402, processor 401 and transceiver 403.
Memory 402 can be independent physical unit, can be connect by bus with processor 401, transceiver 403. Memory 402, processor 401, transceiver 403 also can integrate together, pass through hardware realization etc..
Memory 402 is used to store the program for realizing above method embodiment or Installation practice modules, processing Device 401 calls the program, executes the operation of above method embodiment.
Optionally, pass through when some or all of in the driving behavior classification method based on HMM algorithm of above-described embodiment When software realization, device can also only include processor.Memory for storing program is located at except device, and processor passes through Circuit/electric wire is connect with memory, for reading and executing the program stored in memory.
Processor can be central processing unit (central processing unit, CPU), network processing unit The combination of (network processor, NP) or CPU and NP.
Processor can further include hardware chip.Above-mentioned hardware chip can be specific integrated circuit (application-specific integrated circuit, ASIC), programmable logic device (programmable Logic device, PLD) or combinations thereof.Above-mentioned PLD can be Complex Programmable Logic Devices (complex Programmable logic device, CPLD), field programmable gate array (field-programmable gate Array, FPGA), Universal Array Logic (generic array logic, GAL) or any combination thereof.
Memory may include volatile memory (volatile memory), such as random access memory (random-access memory, RAM)；Memory also may include nonvolatile memory (non-volatile ), such as flash memory (flash memory), hard disk (hard disk drive, HDD) or solid state hard disk memory (solid-state drive, SSD)；Memory can also include the combination of the memory of mentioned kind.
In above-described embodiment, transmission unit or transmitter execute the step of above-mentioned each embodiment of the method is sent, and receive single Member or receiver execute the step of above-mentioned each embodiment of the method receives, and other steps are executed by other modules or processor.Hair Send unit and receiving unit that can form Transmit-Receive Unit, receiver and transmitter can form transceiver.
The embodiment of the present application also provides a kind of computer storage mediums, are stored with computer program, the computer program For executing the driving behavior classification method provided by the above embodiment based on HMM algorithm.
The embodiment of the present application also provides a kind of computer program products comprising instruction, when it runs on computers When, so that computer executes the driving behavior classification method provided by the above embodiment based on HMM algorithm.
It should be understood by those skilled in the art that, embodiments herein can provide as method, system or computer program Product.Therefore, complete hardware embodiment, complete software embodiment or reality combining software and hardware aspects can be used in the application Apply the form of example.Moreover, it wherein includes the computer of computer usable program code that the application, which can be used in one or more, The computer program implemented in usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) produces The form of product.
The application is referring to method, the process of equipment (system) and computer program product according to the embodiment of the present application Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce A raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for real The device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates, Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one The step of function of being specified in a box or multiple boxes.

Claims (10)

1. a kind of driving behavior classification method based on HMM algorithm characterized by comprising
It obtains and drives observation data；
The normal driving likelihood value for driving observation data under normal driving model is calculated to forward algorithm using preceding, And dangerous driving likelihood of the driving behavior data under the dangerous driving model is calculated using the forward algorithm Value；
Bayesian Factor is determined according to the normal driving likelihood value and the dangerous driving likelihood value；
Determining driving behavior classification is compared with threshold value according to the Bayesian Factor.
2. the method according to claim 1, wherein the acquisition drives before observing data, further includes:
Training set is run according to normal driving, normal driving model is established using HMM algorithm, and run and instructed according to dangerous driving Practice collection and dangerous driving model is established using HMM algorithm.
3. according to the method described in claim 2, it is characterized in that, normal driving operation training set and the dangerous driving Data in operation training set include that lateral velocity between adjacent two vehicle is poor, the longitudinal velocity between adjacent two vehicle is poor, adjacent Transverse acceleration between two vehicles is poor, the longitudinal acceleration between adjacent two vehicle is poor, the traffic direction angle between adjacent two vehicle, One of centroid distance between adjacent two vehicle is a variety of.
4. according to claim 1 to method described in 3 any one, which is characterized in that the normal driving model and the danger Dangerous driving model and hidden state sequence, sample observations sequence, state transition probability, state output event random distribution letter Number is related with initial state distribution.
5. according to the method described in claim 4, it is characterized in that, the model posterior probability for driving observation dataI=0 or 1, z be drivings observation data, P (z | λ0) indicate the normal driving likelihood Value, p (z) the expression probability for driving observation data, P (z | λ1) indicate the dangerous driving likelihood value, the Bayes because Son
6. a kind of driving behavior sorter based on HMM algorithm characterized by comprising
Acquiring unit drives observation data for obtaining；
Computing unit observes data normally driving under normal driving model for calculating described drive using forward algorithm Likelihood value is sailed, and danger of the driving behavior data under the dangerous driving model is calculated using the forward algorithm Danger drives likelihood value；
The computing unit, be also used to be determined according to the normal driving likelihood value and the dangerous driving likelihood value Bayes because Son；
Taxon, for being compared determining driving behavior classification with threshold value according to the Bayesian Factor.
7. device according to claim 6, which is characterized in that further include:
Training unit establishes normal driving model using HMM algorithm for running training set according to normal driving, and according to danger Danger drives operation training set and establishes dangerous driving model using HMM algorithm.
8. device according to claim 6 or 7, which is characterized in that the normal driving model and the dangerous driving mould Type and hidden state sequence, sample observations sequence, state transition probability, the probability distribution function of state output event and initial State distribution is related.
9. a kind of sorter of driving behavior, which is characterized in that including processor, input equipment, output equipment and memory, Wherein, for the memory for storing computer program, the computer program includes program instruction, and the processor is for adjusting It is instructed with described program, executes the method according to claim 1 to 5.
10. a kind of computer readable storage medium, which is characterized in that the computer storage medium is stored with computer program, The computer program includes program instruction, and described program instruction makes the processor execute such as right when being executed by a processor It is required that the described in any item methods of 1-5.
CN201811378191.XA 2018-11-19 2018-11-19 Driving behavior classification method and device based on HMM algorithm Pending CN109447182A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811378191.XA CN109447182A (en) 2018-11-19 2018-11-19 Driving behavior classification method and device based on HMM algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811378191.XA CN109447182A (en) 2018-11-19 2018-11-19 Driving behavior classification method and device based on HMM algorithm

Publications (1)

Publication Number Publication Date
CN109447182A true CN109447182A (en) 2019-03-08

Family

ID=65553262

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811378191.XA Pending CN109447182A (en) 2018-11-19 2018-11-19 Driving behavior classification method and device based on HMM algorithm

Country Status (1)

Country Link
CN (1) CN109447182A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110188710A (en) * 2019-06-03 2019-08-30 石家庄铁道大学 Train driver dynamic behaviour recognition methods
CN111666859A (en) * 2020-06-01 2020-09-15 浙江省机电设计研究院有限公司 Dangerous driving behavior identification method
CN111738126A (en) * 2020-06-16 2020-10-02 湖南警察学院 Driver fatigue detection method and device based on Bayesian network and HMM

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104463201A (en) * 2014-11-28 2015-03-25 杭州华为数字技术有限公司 Method and device for recognizing driving state and driver
US20150186714A1 (en) * 2013-12-30 2015-07-02 Alcatel-Lucent Usa Inc. Driver behavior monitoring systems and methods for driver behavior monitoring
CN104835319A (en) * 2015-04-07 2015-08-12 同济大学 Method for estimating vehicle import behavior on high-grade road bottleneck zone on-ramp
CN105528593A (en) * 2016-01-22 2016-04-27 江苏大学 Forward vehicle driver driving behavior prediction system and prediction method
US20170024621A1 (en) * 2015-07-20 2017-01-26 Dura Operating, Llc Communication system for gathering and verifying information
CN107273805A (en) * 2017-05-18 2017-10-20 江苏大学 A kind of GM HMM prediction driving behavior methods of view-based access control model characteristic
CN108693868A (en) * 2018-05-25 2018-10-23 深圳市轱辘车联数据技术有限公司 The method of failure predication model training, the method and device of vehicle trouble prediction

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150186714A1 (en) * 2013-12-30 2015-07-02 Alcatel-Lucent Usa Inc. Driver behavior monitoring systems and methods for driver behavior monitoring
CN104463201A (en) * 2014-11-28 2015-03-25 杭州华为数字技术有限公司 Method and device for recognizing driving state and driver
CN104835319A (en) * 2015-04-07 2015-08-12 同济大学 Method for estimating vehicle import behavior on high-grade road bottleneck zone on-ramp
US20170024621A1 (en) * 2015-07-20 2017-01-26 Dura Operating, Llc Communication system for gathering and verifying information
CN105528593A (en) * 2016-01-22 2016-04-27 江苏大学 Forward vehicle driver driving behavior prediction system and prediction method
CN107273805A (en) * 2017-05-18 2017-10-20 江苏大学 A kind of GM HMM prediction driving behavior methods of view-based access control model characteristic
CN108693868A (en) * 2018-05-25 2018-10-23 深圳市轱辘车联数据技术有限公司 The method of failure predication model training, the method and device of vehicle trouble prediction

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
宋晓琳 等: "基于HMM-SVM的驾驶员换道意图辨识研究*", 《电子测量与仪器学报》 *
熊晓夏 等: "危险换道驾驶行为预测方法研究", 《汽车工程》 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110188710A (en) * 2019-06-03 2019-08-30 石家庄铁道大学 Train driver dynamic behaviour recognition methods
CN110188710B (en) * 2019-06-03 2021-05-04 石家庄铁道大学 Method for identifying dynamic behavior of train driver
CN111666859A (en) * 2020-06-01 2020-09-15 浙江省机电设计研究院有限公司 Dangerous driving behavior identification method
CN111738126A (en) * 2020-06-16 2020-10-02 湖南警察学院 Driver fatigue detection method and device based on Bayesian network and HMM
CN111738126B (en) * 2020-06-16 2023-04-07 湖南警察学院 Driver fatigue detection method and device based on Bayesian network and HMM

Similar Documents

Publication Publication Date Title
Kelly et al. Hg-dagger: Interactive imitation learning with human experts
CN107742193B (en) A kind of driving Risk Forecast Method based on time-varying state transition probability Markov chain
CN106971194B (en) Driving intention recognition method based on improved HMM and SVM double-layer algorithm
CN106740861B (en) The method and apparatus for distinguishing driver based on driving behavior
US9129519B2 (en) System and method for providing driver behavior classification at intersections and validation on large naturalistic data sets
US20210188290A1 (en) Driving model training method, driver identification method, apparatuses, device and medium
Aoude et al. Behavior classification algorithms at intersections and validation using naturalistic data
CN109447182A (en) Driving behavior classification method and device based on HMM algorithm
Hu et al. A framework for probabilistic generic traffic scene prediction
CN104899135A (en) Software defect prediction method and system
JP2022511968A (en) Classifier training to detect open vehicle doors
Hyeon et al. Short-term speed forecasting using vehicle wireless communications
WO2022134711A1 (en) Driving style recognition method, assisted driving method, and apparatuses
Martinsson et al. Clustering vehicle maneuver trajectories using mixtures of hidden markov models
US20220203971A1 (en) Vehicle control method
CN109460065A (en) Unmanned aerial vehicle cluster formation characteristic identification method and system based on potential function
CN107067019A (en) Based on the ADS B under variation Bayesian Estimation and TCAS data fusion methods
Oh et al. In-depth understanding of lane changing interactions for in-vehicle driving assistance systems
Anderson et al. A kinematic model for trajectory prediction in general highway scenarios
Jiang et al. Efficient and unbiased safety test for autonomous driving systems
CN109492597A (en) The method for building up and device of driving behavior model based on SVM algorithm
CN105118332A (en) Air traffic control analog simulation abnormality detection method and device based on clustering analysis method
EP4035090A1 (en) Distance-based learning confidence model
Azevedo et al. Using extreme value theory for the prediction of head-on collisions during passing maneuvres
WO2023230740A1 (en) Abnormal driving behavior identification method and device and vehicle

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
RJ01 Rejection of invention patent application after publication

Application publication date: 20190308

RJ01 Rejection of invention patent application after publication