CN106157657A - Motion state recognition system and method for mobile user - Google Patents

Abstract

The invention discloses a system and a method for identifying the motion state of a mobile user, wherein the method comprises the following steps: the traffic data receiving module is used for collecting the characteristics of the motion state of the mobile user; the traffic data analysis module is used for identifying the motion state of the mobile user so as to distinguish and predict traffic conditions; the analysis result output module is used for recording the motion state of each mobile user and the traffic conditions of different road sections by numerical values; and the display module is used for displaying the real-time traffic conditions of all road sections in different colors on the user terminal and giving out the optimal driving route suggested by the system. By adopting the invention, the motion state of the mobile user can be determined by utilizing a specific algorithm, and the analysis of real-time traffic data is facilitated, so that the utilization rate of the real-time traffic data is improved; and can distinguish and predict traffic patterns and predict traffic conditions of the traffic network more accurately by analyzing the collected mobile traffic data.

Description

A mobile user's motion state recognition system and method thereof

technical field

The invention relates to intelligent transportation and wireless positioning technology, in particular to a mobile user's motion state recognition system and method thereof.

Background technique

In recent years, with the development of society, traffic congestion has become a global problem, which causes a series of consequences while reducing urban efficiency, and is therefore not conducive to urbanization and motorization.

Researchers have been working on solving the traffic congestion problem by providing various types of solutions. One of the solutions is to collect valid traffic information and provide it to drivers to help them make more reasonable driving decisions. Practice has also proved that providing real-time traffic information to public transport participants is a key factor in avoiding road congestion and making optimal driving route decisions.

At present, in mainland China and Hong Kong Special Administrative Region, some private companies and government agencies have begun to provide traffic data to travelers. According to the survey, the transportation departments in the above-mentioned areas have introduced some measures to collect traffic data, and can use different methods to receive traffic data of different types of vehicles, but there is still a lack of technology for analyzing traffic data. Since traffic data usually exists in digital form, drivers cannot use it directly. According to the results of the survey, the utilization rate of traffic data in developed countries or regions in the world is still very low.

Therefore, it is urgent to study a technology that can fully mine and utilize traffic data to improve the utilization rate of real-time traffic data.

Contents of the invention

In view of this, the main purpose of the present invention is to provide a mobile user's motion state recognition system and method thereof, which can determine the motion state of the mobile user through a specific algorithm, help analyze real-time traffic data, and improve the utilization rate of real-time traffic data.

Another object of the present invention is to identify and predict traffic patterns and more accurately predict traffic conditions of the traffic network by analyzing collected mobile traffic data.

In order to achieve the above object, technical solution of the present invention is achieved in that way:

A motion state recognition system for mobile users, comprising a traffic data collection module, a traffic data analysis module, an analysis result output module, and a display module; wherein:

The traffic data receiving module is used to collect the characteristics of the motion state of the mobile user;

The traffic data analysis module is used to identify the motion state of the mobile user, so as to identify and predict traffic conditions;

The analysis result output module is used to numerically record the motion state of each mobile user and the traffic conditions of different road sections; and,

The display module is used to display the real-time traffic conditions of each road section in different colors on the user terminal, and provide the optimal driving route suggested by the system.

Wherein: the characteristics of the motion state of the mobile user include speed, position, time, driving direction, and braking information.

The motion state of the mobile user includes a static state, a walking state, and a state of moving with a motor vehicle.

The state of moving with the motor vehicle specifically includes two states of a low-speed motor vehicle and a high-speed motor vehicle.

A method for identifying a motion state of a mobile user, comprising the steps of:

A. Steps for collecting mobile traffic data;

B. Use the data analysis module with built-in support vector machine to analyze the mobile traffic data, and distinguish the steps of various mobile user categories and motion states;

C. Judging whether the moving speed of the mobile user within a unit time is greater than a certain preset speed value, if yes, then judging that he is in a driving car; if not, judging that he is not in a driving car.

Wherein: step C further includes: D, the step of switching the application mode of the mobile user terminal as required, specifically: closing the data analysis mode or opening the data analysis mode.

Step D further includes: if the data analysis mode is turned off, maintaining the current traffic state; otherwise, returning to step A.

The process of collecting mobile traffic data is: using various mobile terminals or/and vehicles connected to the wireless communication network to collect user information and motion state information of mobile users.

The traffic data analysis module of the support vector machine is built in the mobile terminal and the vehicle.

The vehicle accesses the V2V network.

The user information includes terminal and vehicle type data for collecting mobile traffic data.

The motion status information includes motion speed, location, time, driving direction, braking and weather condition information.

The motion state recognition system and method of the mobile user provided by the present invention have the following advantages:

As mentioned above, identifying the motion status of mobile users, such as stationary, walking, and in low-speed high-speed vehicles, is very important for collecting real-time traffic information using mobile devices. By using this algorithm, analysts can easily select appropriate traffic data for further analysis. Therefore, this invention can optimize traffic data utilization.

Compared with other machine learning algorithms, the mobile user's motion state recognition algorithm further improves the classification accuracy rate of the algorithm because it maps low-dimensional data to high-dimensional space, and the accuracy rate can reach 94.93% in the actual measurement of the project. When training the motion state recognition algorithm model of the mobile user, it can still achieve a high accuracy rate when there are few training samples, which makes the algorithm more widely and quickly used without waiting for a long time to collect data. Compared with other machine learning algorithms, such as neural network algorithms, the mobile user's motion state recognition algorithm has a small amount of calculation, requires less calculation time, and has lower requirements on memory and hardware.

Description of drawings

Fig. 1 is the optimal classification plane schematic diagram of the support vector machine (SVM) that adopts in the motion state recognition system of mobile user in the embodiment of the present invention;

FIG. 2 is a schematic diagram of the composition structure of a mobile user's motion state recognition system according to an embodiment of the present invention;

FIG. 3 is a schematic flowchart of a method for identifying a mobile user's exercise state according to an embodiment of the present invention.

detailed description

The mobile user's exercise state recognition system and method thereof of the present invention will be further described in detail in conjunction with the accompanying drawings and the embodiments of the present invention.

The present invention identifies and predicts traffic patterns by analyzing collected mobile traffic data, combined with data collected from other sources, such as coils, GPS positioning vehicles, V2V networks, and video surveillance. These results will be used to predict some road sections that cannot collect real-time traffic data or predict the future road conditions of some road sections that can provide real-time traffic data. According to the Bayesian network method can be good at dealing with traffic information from adjacent roads, and can also deal with incomplete data. Therefore, we can more accurately predict the traffic conditions of the traffic network based on incomplete data.

Here, the V2V network is a mesh network. Nodes in the network (such as cars, smart traffic lights, etc.) can emit, capture and forward signals. On this network, cars send messages to each other to tell each other What you are doing, the information includes but not limited to speed, location, driving direction, braking, etc. The V2V technology uses Dedicated Short-Range Communication (DSRC), which has a coverage range of up to 300 meters, and it can also use the hops of multiple nodes on the network to collect traffic conditions 1.5 kilometers away, which is enough for most drivers coping time.

The present invention is used to identify the motion state of a mobile user, such as standing still, walking, in a low-speed or high-speed vehicle, etc., which is particularly important for using mobile devices to collect real-time traffic information. In order to accurately predict the user's motion state, we use the Extended Class Association Rules (CAR) algorithm to establish some valuable rules. These rules come from covering various types of real user information (such as speed, location, time, driving direction, braking, etc.), and can be easily integrated with important domain knowledge. The proposed method, by applying these rules and considering some uncertainties, can meet the requirements of the dynamic and open environment of Intelligent Transportation Systems (ITS). And by identifying the state of motion of the mobile user, such as stationary, walking, in low-speed, high-speed vehicles, etc., the method of the present invention can easily select suitable traffic data to carry out the next step of analysis, thus can be used to improve traffic conditions. Data utilization.

As we all know, the activity state of mobile users can be divided into static state, walking state and state of random motor vehicle movement (mobile users who use non-motorized vehicles to travel generally maintain a static state or a walking state). Therefore, the acquired mobile user samples can be divided into pedestrian and motor vehicle categories. For mobile users who move randomly by motor vehicle, there may be a static state during the motion process, and the static state may have a certain impact on the final calculated traffic information. The reasons for the static state generally include the following: parking at the destination, waiting time for changing cars, artificial parking, and parking caused by traffic jams.

There is a mobile user in a "stationary" state generated by the first three situations during the movement (classified as motor vehicle interference), regardless of the traffic conditions of the road section where it is located, and the length of the static time, its speed will be lower than that of the same road section There is no "stationary" state on the speed of mobile users moving with the car (classifying it as a motor vehicle); while the mobile users in the "stationary" state generated in the fourth case are generated by the actual traffic conditions on the road surface, It will not affect the final calculation result.

As mentioned above, the moving speed of the mobile users includes: the speed of the walking mobile users, the speed of the motor vehicle interference mobile users, and the speed of the motor vehicle mobile users.

In order to calculate the average speed information of each road section in the road network, the present invention utilizes the method of SVM (Support Vector Machines) classification to extract the speed information of the mobile users of motor vehicles from the speed information of the above mobile users. From the previous analysis, we can see that the average speed of mobile users with motor vehicle interference is always not greater than the average speed of mobile users with motor vehicles, while the average speed of mobile users with pedestrians is slightly affected by traffic conditions and is stable at about 3 ~5km/h (typical value can be 4km/h).

The present invention uses the method of support vector machine (SVM) classification to divide the speed of mobile users into 3 classes, then judges the class of each class according to the above-mentioned characteristics, the class whose centroid is closest to 4km/h is the walking class, and the remaining 2 classes Among the classes, the one with the smaller center of mass is the motor vehicle interference class, and the one with the larger center of mass is the motor vehicle class. Finally, according to the speed of mobile users of motor vehicles, the average speed or average travel time of each road section in the road network can be calculated.

The invention adopts a support vector machine (SVM) to classify to identify the motion state of the mobile user.

Here, the machine learning method we are using is Support Vector Machine (SVM). The support vector machine has many unique advantages in solving small sample, nonlinear and high-dimensional pattern recognition, and can be extended to other machine learning problems such as function fitting.

The support vector machine (SVM) method is based on the VC dimension theory of statistical learning theory and the principle of structural risk minimization. According to the complexity of the model (that is, the learning accuracy of specific training samples, Accuracy) and learning Ability (i.e., the ability to identify arbitrary samples without error) to seek the best balance, in order to obtain the best generalization ability.

FIG. 1 is a schematic diagram of an optimal classification surface of a support vector machine (SVM) used in a mobile user's motion state recognition system according to an embodiment of the present invention.

The support vector machine (SVM) is developed from the optimal classification surface in the case of linear separability, and its basic idea can be illustrated by the two-dimensional case shown in the figure below.

As shown in Figure 1, C ₁ and C ₂ respectively represent the two types of data samples to be distinguished; H represents the classification line _y = ωx + b; H ₁ and H ₂ are parallel to H and pass the two types of samples closest to H The straight line; the distance between H ₁ and H, H ₂ and H is called the geometric interval, and its expression is:

$\mathrm{<span\; class="notranslate">\; \&delta;</span>}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; \&omega;x</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; |</span>}{<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; \&omega;</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>}<span\; class="notranslate">\; ;</span>$

In the above formula, ||ω|| represents the norm of ω; b represents the classification penalty.

The so-called optimal classification line means that it can not only correctly separate the two types of data samples, minimize the training error rate, but also maximize the geometric interval. The former guarantees the minimum empirical risk; while the latter actually minimizes the confidence range in the generalization bound, and the larger the geometric interval, the smaller the upper bound of the error. Expanding from two-dimensional space to high-level space, the optimal classification line becomes the optimal classification surface. The corresponding classification problem can be transformed into a constrained minimization problem. We detail it in the next section.

Wherein, in the motion state recognition system of mobile user of the present invention, the Support Vector Machine (SVM) algorithm that adopts, its description is as follows:

Assume that there are l training samples, represented by vector _xi ∈ R ⁿ , i=1,2,...,l. The training sample is the user's mobile data in this project. The category of the final classification takes two categories as an example, represented by |y _i {1,-1}, which can represent the user's static or mobile state. For this classification problem, the following mathematical expression is used to calculate.

$\mathrm{<span\; class="notranslate">\; min</span>}_{\mathrm{<span\; class="notranslate">\; w</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; e</span>}}<span\; class="notranslate">\; (</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; w</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; C</span>}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>$

subject to:

ε _i ≥ 0, 1=1, . . . , 1.

In formula (1), _εi represents the relaxation factor, which allows the existence of misclassified samples; C is a positive constant, called the penalty factor; As a penalty item, the purpose of introducing it is to seek a certain balance between empirical risk and generalization performance. is to project _xi into a higher dimensional space. Since the vector parameter w may have a high dimensionality, and this problem is an optimization problem with inequality constraints, it can be solved by adding Lagrange multipliers and constructing Lagrange functions, and finally transforming the problem of solving the optimal classification surface above into The dual problem for optimization of the following convex quadratic programming:

$\underset{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}{\mathrm{<span\; class="notranslate">\; min</span>}}\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; Q\&alpha;</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

Subject to: y ^T α=0, 0≤α _i ≤C, 1=1...1,

In formula (2), e=[1,...,] ^T is a unit vector, Q is a positive semi-definite square matrix of l×l, α _i is the corresponding Lagrange multiplier, Q _if ≡y _i y _f K(x _i , x _f . It is a kernel function, which can realize the mapping from low-dimensional space to high-dimensional space, so as to solve nonlinear problems. Commonly used kernel functions include polynomial function, radial basis function and Sigmoid function.

In the present invention, the kernel function adopted is radial basis function:

exp(-δr ² ),

where r=1/n.

By solving model (2) and using the correspondence between the dual model and the original model, the vector parameter w in model (1) is calculated to satisfy the following conditions:

The final calculated optimal classification surface function is:

This is the final decision function. For a new test sample, if the value of the optimal classification surface function is a negative value, the test sample is classified into the category of y=-1, and if the value of the optimal classification surface function is a positive value, the test sample Samples are assigned to the class y=1. Corresponding to this project, we first collect the actual mobile data of the user as a training sample, and use the calculation of the model (2) to obtain the optimal classification surface in the final (4).

In combination with the technical solution of the present invention, y represents the category to which the mobile user belongs, that is, static and mobile, and _xi represents the characteristics of the mobile user's motion state, such as speed, position, time, driving direction, braking, etc.

By judging whether the value of the optimal classification surface function is positive or negative, the users are divided into two categories: stationary and mobile. At the same time, for the category of moving state, it can also be divided into walking category and motor vehicle category through the above algorithm. At this time, y represents these two categories, and _xi represents the characteristics of the mobile user's motion state, such as speed, position, time, driving direction, braking, etc.

FIG. 2 is a schematic diagram of the structure of the mobile user's exercise state recognition system according to the embodiment of the present invention. As shown in Figure 2, the motion state recognition system of the mobile user mainly includes a traffic data collection module, a traffic data analysis module, an analysis result output module and a display module. in:

The traffic data receiving module is used to collect the characteristics of the mobile user's motion state, including speed, position, time, driving direction and braking, etc.;

The traffic data analysis module is used to identify the motion state of the mobile user, including a static state, a walking state, and a state of random motor vehicle motion (including two states of a low-speed motor vehicle and a high-speed motor vehicle); predict traffic conditions;

The analysis result output module is used to record the motion state of each mobile user and the traffic conditions of different road sections with numerical values;

The display module is used to display the real-time traffic conditions of each road section in different colors on the user terminal, and provide the optimal driving route suggested by the system.

FIG. 3 is a schematic flowchart of a method for identifying a mobile user's exercise state according to an embodiment of the present invention. As shown in Figure 3, the motion state recognition method of described mobile user comprises the steps:

Step 31: Step for collecting mobile traffic data.

Here, various mobile terminals and/or equipment such as vehicles or vehicles can be used to collect user information and exercise state information of mobile users. The mobile terminal includes, but is not limited to, an intelligent mobile communication terminal, a GPS terminal, and a tablet computer capable of wirelessly accessing a communication network. The vehicle refers to various means of transportation connected to the V2V network, and usually specifically refers to various motor vehicles, such as cars, buses, freight vehicles, and the like. The user information includes, but is not limited to, data such as terminals and vehicle types that collect mobile traffic data. The movement status information includes but not limited to movement speed, location, time, driving direction, braking and weather conditions and other information.

Step 32: Using a traffic data analysis module with a built-in support vector machine (SVM) to analyze the mobile traffic data to distinguish various mobile user categories and motion states; then execute step 33.

Here, the specific process of using the traffic data analysis module to analyze the mobile traffic data is as follows:

Step 33: Judging whether the moving speed of the mobile user within a unit time is greater than a certain preset speed value, if yes, execute step 34; if not, execute step 35.

Step 34: Determine that it is on a moving vehicle, and then execute Step 36.

Step 35: Determine that it is not on a moving vehicle, and then execute Step 36.

Step 36: The step of switching the application mode of the mobile user terminal as required, and then performing step 37.

Step 37: Determine whether the application mode of the mobile user terminal has changed, if not, return to step 31, and continue to analyze the collected mobile traffic data; if so, execute step 38, and exit the data analysis mode.

Here, whether the application mode of the mobile user terminal is changed specifically includes: closing the data analysis mode or opening the data analysis mode.

Step 38: Maintain the current traffic state.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention.

Claims (12)

1. the moving state identification system of a mobile subscriber, it is characterised in that comprise traffic data collection module, traffic data analyzing module and analysis result output module and display module；Wherein:

Described traffic data receiver module, for collecting the feature of mobile subscriber's motion state；

Described traffic data analyzing module, for identifying the motion state of mobile subscriber, distinguishes with this and predicts traffic；

Described analysis result output module, is used for the traffic of the motion state with each mobile subscriber of records of values and different sections of highway；And,

Described display module, for showing the real-time traffic condition in each section on the subscriber terminal in different colors, provides the optimum drive route of system recommendations.

2. the moving state identification system of mobile subscriber according to claim 1, it is characterised in that the feature of described mobile subscriber's motion state, including speed, position, time, travel direction, brake information.

3. the moving state identification system of mobile subscriber according to claim 1, it is characterised in that the motion state of described mobile subscriber, including inactive state, ambulatory status and the state with motor vehicle motion.

4. the moving state identification system of mobile subscriber according to claim 3, it is characterised in that the described state with motor vehicle motion is specifically included in low-speed machine motor-car and High speed vehicle two states.

5. the moving state identification method of a mobile subscriber, it is characterised in that comprise the steps:

A, the step for collecting mobile traffic data；

B, utilize and be built-in with the data analysis module of SVMs mobile traffic data are analyzed, the step differentiating various mobile subscriber's classification and motion state；

C, judge that whether translational speed within the unit interval for the described mobile subscriber is more than a certain pre-set velocity value, if so, then judge it on the car travelling；If it is not, then judge it not on the car travelling.

6. the moving state identification method of mobile subscriber according to claim 5, it is characterised in that farther include after step C:

D, the step changing mobile subscriber terminal application model as required, particularly as follows: close data analysis pattern or turn-on data analytical model.

7. the moving state identification method of mobile subscriber according to claim 6, it is characterised in that step D farther includes:

If closing data analysis pattern, then keep current traffic condition；Otherwise, step A is returned.

8. the moving state identification method of mobile subscriber according to claim 5, it is characterized in that, the process collecting mobile traffic data is: utilize the mobile terminal of various access to wireless communication network or/and the vehicles are collected user profile and the movement state information of mobile subscriber.

9. the moving state identification method of the mobile subscriber according to claim 5 or 8, it is characterised in that be built-in with the traffic data analyzing module of SVMs in described mobile terminal and the vehicles.

10. the moving state identification method of the mobile subscriber according to claim 5 or 8, it is characterised in that the described vehicles access V2V network.

The moving state identification method of 11. mobile subscribers according to claim 8, it is characterised in that described user profile, including collect terminal and the type of vehicle data of mobile traffic data.

The moving state identification method of 12. mobile subscribers according to claim 8, it is characterised in that described movement state information, including movement velocity, position, time, travel direction, brake and weather conditions information.