CN110781961B - Accurate behavior recognition method based on decision tree classification algorithm - Google Patents
Accurate behavior recognition method based on decision tree classification algorithm Download PDFInfo
- Publication number
- CN110781961B CN110781961B CN201911025926.5A CN201911025926A CN110781961B CN 110781961 B CN110781961 B CN 110781961B CN 201911025926 A CN201911025926 A CN 201911025926A CN 110781961 B CN110781961 B CN 110781961B
- Authority
- CN
- China
- Prior art keywords
- behavior recognition
- information
- data
- base station
- user
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000003066 decision tree Methods 0.000 title claims abstract description 23
- 238000007635 classification algorithm Methods 0.000 title claims abstract description 10
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 36
- 238000002474 experimental method Methods 0.000 claims abstract description 11
- 238000013480 data collection Methods 0.000 claims abstract description 7
- 238000000638 solvent extraction Methods 0.000 claims abstract description 4
- 230000006399 behavior Effects 0.000 claims description 64
- 238000012549 training Methods 0.000 claims description 21
- 210000000629 knee joint Anatomy 0.000 claims description 12
- 238000005192 partition Methods 0.000 claims description 9
- 210000003127 knee Anatomy 0.000 claims description 8
- 230000000694 effects Effects 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 4
- 238000012360 testing method Methods 0.000 claims description 4
- 238000006073 displacement reaction Methods 0.000 claims description 3
- 239000000203 mixture Substances 0.000 claims description 3
- 230000009897 systematic effect Effects 0.000 claims description 3
- 238000005303 weighing Methods 0.000 claims description 3
- 238000012795 verification Methods 0.000 abstract description 3
- 238000005516 engineering process Methods 0.000 description 8
- 238000007637 random forest analysis Methods 0.000 description 6
- 238000011161 development Methods 0.000 description 4
- 230000018109 developmental process Effects 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 230000003542 behavioural effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 230000035800 maturation Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 230000000284 resting effect Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/02—Services making use of location information
- H04W4/029—Location-based management or tracking services
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Signal Processing (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention discloses an accurate behavior recognition method based on a decision tree classification algorithm, which comprises the following steps: s1, data collection: solving the label position by adopting a Chan algorithm based on TDOA; s2, extracting features: selecting characteristic values and partitioning characteristic value intervals; s3, behavior recognition: and establishing a user behavior recognition model. According to the method for realizing behavior recognition based on the decision tree algorithm, firstly, the sensor is used for acquiring user data, then the extracted characteristic value is recognized, and finally, behavior classification experiment verification is realized by adopting the decision tree algorithm. Experimental results show that the algorithm has good daily behavior recognition performance under specific conditions.
Description
Technical Field
The invention relates to a behavior recognition method, in particular to an accurate behavior recognition method based on a decision tree classification algorithm.
Background
Currently, with the continuous progress of wireless technologies such as bluetooth Zigbee UWB (ultra wideband), wireless network and the like, the explosion of wireless sensor networks for ad hoc networks and internet of things has attracted extensive attention in academia. Behavior recognition is widely recognized as an indispensable key technology. The behavior identification information facilitates pre-warning in network emergency situations, decision-making during events, and post-processing of incidents. Behavior recognition technology plays a decisive role in the further development of wireless networks. Therefore, research into behavior recognition technology is particularly important.
In indoor locations, it is difficult to accurately analyze parameters such as signal arrival time or arrival angle because the signal propagation environment is more complex than that of outdoor. However, with the continued advancement and development of wireless sensor networks WSNs, academic research has not been limited to traditional indoor positioning and location awareness. Currently, many fields have begun to utilize radio for position sensing, among which rapid development and maturation of position sensing technology based on UWB (ultra wide band) radar systems are most prominent. In the reference, a UWB channel based on human body shielding is provided, and the influence of the human body shielding on TOA ranging error is studied by measuring and analyzing the ranging error of TOA. At present, the novel indoor positioning technology based on commercial Wi-Fi equipment has good development advantages in all aspects. Such as indoor intrusion detection, campus security, market personnel detection, patient monitoring, and real-time detection of elderly people and children in the home.
The frequency band of the 2.4GHz wireless network is similar to Bluetooth, and the positioning method is also affected by the environment and inaccurate in data when encountering obstacles or electromagnetic interference. Compared with the characteristics of various positioning technologies, UWB is a long-term research hotspot in the field of domestic and foreign radio frequency communication. Many existing behavior awareness methods employ image processing methods. And extracting low-level features by utilizing image information, identifying human body movement and constructing a human body movement mode. But the disadvantage is that the extraction amount of the characteristic value is large, and the safety and privacy of the user are seriously threatened. Therefore, more and more technologies replace image processing methods with sensors that are perceptively small in size, inexpensive to deploy, and simple to tamper with. The existing algorithm adopts big data parallel classification to guide the power supply mode, but does not consider the energy consumption problem; some methods refine the original classification data and propose the maximum attribute index algorithm in the same-level conceptual refinement. The privacy budget is more reasonably allocated between different levels using a hierarchical geometric allocation mechanism. However, data distribution cannot be implemented in a dynamic data environment. Some methods improve the objective function of the decision tree generation algorithm, so that inconsistent data can be classified, and the influence factors of the function can be directly adjusted, so that the node segmentation of the decision tree is more accurate, and the classification effect is better. Some methods use a compression strategy selected based on HBase data classification. However, the data processing procedure is relatively complex.
Disclosure of Invention
The invention mainly aims to provide an accurate behavior recognition method based on a decision tree classification algorithm.
The technical scheme adopted by the invention is as follows: an accurate behavior recognition method based on a decision tree classification algorithm comprises the following steps:
s1, data collection: solving the label position by adopting a Chan algorithm based on TDOA;
s2, extracting features: selecting characteristic values and partitioning characteristic value intervals;
s3, behavior recognition: and establishing a user behavior recognition model.
Further, the step S1 specifically includes:
considering positioning accuracy and equipment cost, 4 base stations are selected more appropriately; in the two-dimensional plane rectangular coordinate system, the coordinate of the ith base station is B uwb,i ＝[x i ,y i ] T (i=1, 2, …, 5) the coordinates of the tag are T uwb ＝[x 0 ,y 0 ] T A non-line of sight between the base station and the tag is R i ＝||B uwb,i -T uwb || 2 (i=1, 2, …, 5); taking a first base station as a common reference node to obtain a group of TDOA observation values delta t i,1 (i=2, 3,4, 5) indicating a signal arrival time difference between the i-th base station and the first base station;
in the case of this model of the present invention,is Deltat i,1 True value of n i,1 Measured by systematic error, NLOS error is n NLOS,i ；
Let the signal propagation velocity be c and calculate R i,1 Difference between the distances from the mark to the i-th base station and the first base station:
R i,1 ＝c·Δt i,1 (i＝2,3,4,5) (2)
establishing three hyperbolic equations R according to hyperbolic characteristics i,1 ＝R i -R 1 (i=2, 3,4, 5), regarding T uwb Can be established as shown in formula (3);
adopting a label positioning architecture of 4 base stations, taking one base station as a main part, and the rest 3 base stations; when a tester carrying a locating tag enters a test area, signals sent by the tag are received by one or more sensors; decoding signals from the sensors that transmit angle of arrival and timing information, and then transmitting these data to the master sensor; the main sensor collects all information sent by the base station to calculate the position information of the tag, so that data collection is realized; the sensor transmits data every second through the switch and the server, the data adopts a UDP data packet format, and the server receives the UDP data packet and can acquire the X, Y coordinate information specific to the tag.
Further, the step S2 specifically includes:
the feature value selection includes: position division, head-shoulder waist-knee height processing and distance movement in unit time of head-shoulder waist and knee joint;
the location division includes:
in real life, the position of a user has a certain relation with the behavior activity of the user; spatial locations fall into three categories: first, the user can sit in a place where he lies for rest; the second type is a distance of 0.1 to 0.3 meters from the region, denoted Da, depending on the object; the remaining space is of the third class, denoted La;
the head, shoulder, waist and knee height treatment comprises head height, shoulder height, waist height and knee height;
z-axis data of the head, shoulder, waist and knee joints of a user represent the height of a user space and are directly read from tag coordinates;
the distance moves within unit time of the head, shoulder, waist and knee joints and comprises the distance from the head to the shoulder to the waistline and the knee;
direct calculation of the distance between the head, shoulder and waist of the user and the knee joint in unit time is difficult to achieve, mainly because the unit time is difficult to determine; the excessive numerical value influences the displacement calculation result, so that the user behavior cannot be accurately described, and the accuracy of the user behavior recognition error is reduced; too small a value increases the amount of computation due to delay overhead; through multiple experiments, the optimal unit time of weighing accuracy and operation delay is LS;
the partition characteristic value interval comprises:
after determining the above-mentioned feature values, classification boundaries must be determined to ensure similarity between data and intra-category, inter-category differences; combining experiments, and processing the mixture by adopting a layering classification method; classification focuses on how to determine the boundaries of each level; currently, determining boundary values is two algorithms: length equivalent method and distributed equivalent method;
set the range of eigenvalues phi= [ c ] min ,c max ]The method is divided into N grades, and grade labels are 1-N; from the value range phi= [ c min ,c max ]A range of sensor values r=c can be obtained max -c min The method comprises the steps of carrying out a first treatment on the surface of the To ensure that the length of each interval within the region is the same; then calculating to obtain the length of each interval as r=r/N; thus, the value range of each section can be determined.
Further, the step S3 specifically includes:
let D be the training tuples divided by category, then the entropy of D is expressed as:
wherein p is i Representing the probability that the ith class appears in the entire training tuple, can be estimated by dividing the number of elements belonging to that class by the total number of elements in the training tuple; the actual meaning of entropy represents the average amount of information required for tuple class labels in D;
let the training tuple D divided by the attribute A, the expected information for the D partition is:
the information delta is the difference between them:
gain(A)＝inf o(D)-inf o A (D) (6)
establishing a user behavior recognition model; the specific behavior recognition steps are as follows:
s31, classifying and collecting various behaviors according to a training data set, and dividing a training tuple into entropy info (D) of a training tuple;
s32, extracting position height characteristic values from data preprocessing, calculating characteristic value intervals and dividing the characteristic values;
s33, obtaining information gain (A) about the characteristic value of the partition before the prediction in the step 2;
s34, the difference between the information increment is caused by the expected information difference;
when the maximum value of the output is reached, memorizing the maximum gain as the maximum memory; the delta information is then the corresponding behavior information.
The invention has the advantages that:
according to the method for realizing behavior recognition based on the decision tree algorithm, firstly, the sensor is used for acquiring user data, then the extracted characteristic value is recognized, and finally, behavior classification experiment verification is realized by adopting the decision tree algorithm. Experimental results show that the algorithm has good daily behavior recognition performance under specific conditions.
In addition to the objects, features and advantages described above, the present invention has other objects, features and advantages. The present invention will be described in further detail with reference to the drawings.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention.
FIG. 1 is a diagram of a behavior recognition framework of the present invention;
FIG. 2 is a user body part compartmentalization of the wait for test of the present invention;
FIG. 3 is a schematic representation of an experimental environment of the present invention;
fig. 4 is a comparison graph of behavior recognition error rates of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
Referring to fig. 1, as shown in fig. 1, a method for identifying accurate behaviors based on a decision tree classification algorithm includes the following steps:
s1, data collection: solving the label position by adopting a Chan algorithm based on TDOA;
s2, extracting features: selecting characteristic values and partitioning characteristic value intervals;
s3, behavior recognition: and establishing a user behavior recognition model.
The step S1 specifically includes:
considering positioning accuracy and equipment cost, 4 base stations are selected more appropriately; in the two-dimensional plane rectangular coordinate system, the coordinate of the ith base station is B uwb,i ＝[x i ,y i ] T (i=1, 2, …, 5) the coordinates of the tag are T uwb ＝[x 0 ,y 0 ] T A non-line of sight between the base station and the tag is R i ＝||B uwb,i -T uwb || 2 (i=1, 2, …, 5); taking a first base station as a common reference node to obtain a group of TDOA observation values delta t i,1 (i=2, 3,4, 5) indicating a signal arrival time difference between the i-th base station and the first base station;
in the case of this model of the present invention,is Deltat i,1 True value of n i,1 Measured by systematic error, NLOS error is n NLOS,i ；
Let the signal propagation velocity be c and calculate R i,1 Difference between the distances from the mark to the i-th base station and the first base station:
R i,1 ＝c·Δt i,1 (i＝2,3,4,5) (2)
three hyperbolas are established according to hyperbolic characteristicsEquation R i,1 ＝R i -R 1 (i=2, 3,4, 5), regarding T uwb Can be established as shown in formula (3).
Adopting a label positioning architecture of 4 base stations, taking one base station as a main part, and the rest 3 base stations; when a tester carrying a locating tag enters a test area, signals sent by the tag are received by one or more sensors; decoding signals from the sensors that transmit angle of arrival and timing information, and then transmitting these data to the master sensor; the main sensor collects all information sent by the base station to calculate the position information of the tag, so that data collection is realized; the sensor transmits data every second through the switch and the server, the data adopts a UDP data packet format, and the server receives the UDP data packet and can acquire the X, Y coordinate information specific to the tag.
The step S2 specifically includes:
the feature value selection includes: position division, head-shoulder waist-knee height processing and distance movement in unit time of head-shoulder waist and knee joint;
the location division includes:
in real life, the position of a user has a certain relation with the behavior activity of the user; for example, on a sofa, a user may sit or lie down; in hallways, when a user falls or lies near an item at home, they may run while walking, which is easier to fall than if they were near an item at home; thus, spatial locations fall into three categories: first, the user can sit in a resting area, such as a sofa, a bed, a chair, etc., recorded as Ra (reserra); the second type is a distance of 0.1 to 0.3 meters from the region, denoted Da (distanceaera), depending on the object; the remaining space is of the third class, denoted La (Last aero); the division for measuring the body parts of the user is shown in fig. 2;
the head-shoulder-waist-knee height treatment includes (head height, shoulder height, waist height, knee height);
z-axis data of the head, shoulder, waist and knee joints of a user represent the height of a user space and can be directly read from tag coordinates;
the distance moves within a unit time of the head-shoulder waist and knee joint includes (the distance of the head from the shoulder from the waistline knee);
direct calculation of the distance between the head, shoulder and waist of the user and the knee joint in unit time is difficult to achieve, mainly because the unit time is difficult to determine; the excessive numerical value influences the displacement calculation result, so that the user behavior cannot be accurately described, and the accuracy of the user behavior recognition error is reduced; too small a value increases the amount of computation due to delay overhead; through multiple experiments, the optimal unit time of weighing accuracy and operation delay is LS;
the partition characteristic value interval comprises:
after determining the above-mentioned feature values, classification boundaries must be determined to ensure similarity between data and intra-category, inter-category differences; combining experiments, and processing the mixture by adopting a layering classification method; classification focuses on how to determine the boundaries of each level; currently, determining boundary values is two algorithms: length equivalent method and distributed equivalent method;
set the range of eigenvalues phi= [ c ] min ,c max ]The method is divided into N grades, and grade labels are 1-N; from the value range phi= [ c min ,c max ]A range of sensor values r=c can be obtained max -c min The method comprises the steps of carrying out a first treatment on the surface of the To ensure that the length of each interval within the region is the same; then calculating to obtain the length of each interval as r=r/N; thus, the value range of each section can be determined. For example, the value range of the i-th section is [ c min +(i-1)r,c min +ir]。
The step S3 specifically includes:
the decision tree algorithm is a method for approximating discrete function values, which is a typical classification method, and the basic idea is to process data firstly, then generate readable rules and decision trees by using a generalization algorithm, and then analyze new data by using a decision; essentially, a decision tree is a process of classifying data by a series of rules; common decision tree classification algorithms include ID3, C4.5, CART, etc.; the smaller the expected information, the larger the information gain and the higher the purity, while the core of the ID3 algorithm thinks to choose the information gain as an attribute; the ID3 algorithm is used for identification. Let D be the training tuples divided by category, then the entropy of D is expressed as:
wherein p is i Representing the probability that the ith class appears in the entire training tuple, can be estimated by dividing the number of elements belonging to that class by the total number of elements in the training tuple; the actual meaning of entropy represents the average amount of information required for tuple class labels in D;
now suppose that training tuple D is divided by attribute A, then the expected information for the D partition is:
the information delta is the difference between them:
gain(A)＝inf o(D)-inf o A (D) (6)
the ID3 algorithm calculates the gain rate of each attribute every time the division is needed, and then selects the attribute with the maximum gain rate for division; therefore, as long as the maximum gain ratio can be found, the optimal dividing effect can be obtained;
according to the analysis, the abstract description of the user behavior is a behavior classification model and is also the basis of behavior recognition; based on the decision tree classification algorithm, a user behavior recognition model is established; the specific behavior recognition steps are as follows:
s31, classifying and collecting various behaviors according to a training data set, and dividing a training tuple into entropy info (D) of a training tuple;
s32, extracting position height characteristic values from data preprocessing, calculating characteristic value intervals and dividing the characteristic values;
s33, obtaining information gain (A) about the characteristic value of the partition before the prediction in the step 2;
s34, the difference between the information increment is caused by the expected information difference;
when the maximum value of the output is reached, memorizing the maximum gain as the maximum memory; the delta information is then the corresponding behavior information.
To verify the performance of the algorithm in behavior recognition, a scenario experiment was performed and three scenarios were selected from the office conference room and laboratory. The resulting behavior recognition accuracy is compared to other algorithms. In the experiment, researchers held the positioning tag, simulating six basic actions of sitting, standing, falling, lying, walking and running. The experimental environment plan is shown in fig. 3:
during the measurement, there may be many data different from the actual behavior to be eliminated during the preprocessing due to various behavioral instabilities simulated by the human hand tag and instability of the sensor itself. Finally, as shown in table 1, valid data is provided.
TABLE 1 effective dataset
The region indicates that three features of sitting, lying and walking are three features respectively. Y stands for user behavior, 1 for sitting, 2 for standing, 3 for falling, 4 for lying, 5 for walking, 6 for running. The units of height and distance are meters.
Through algorithm analysis, the accuracy of each behavior recognition is found to be classified into three types. All the characteristic values participate in class 1, and the identification result is shown in table 2; the second category does not contain the location feature recognition results, which are shown in table 3; the third class does not contain high feature recognition results, which are shown in table 4.
TABLE 2 behavior correctness of all eigenvalues (%)
It can be seen from table 2 that both standing and lying behaviors are most easily identified, since both states are more easily identified and the tag data features are obvious, i.e. there is no significant change in height and position. Namely, the moving distance per unit time is 0; the recognition effects of sitting and walking states are relatively different, and the movement positions and heights of the people are slightly changed and the characteristic values are relatively close; because the range of variation is also better distinguished.
TABLE 3 accuracy of behavior recognition except for position eigenvalues (%)
As can be seen from table 3, the recognition rate of each behavior is lower than that of all the eigenvalues; standing and lying behaviors are still the easiest to identify; the standing and sitting posture recognition effect is not obvious and is influenced by the position characteristic value; however, knowledge of the decline and the operating state remains apparent.
TABLE 4 accuracy of behavior recognition except for height eigenvalues (%)
In table 4, it is shown that standing and lying behaviors are most easily recognized because both states are less affected by the height characteristic value, and walking and sitting recognition rates are significantly reduced.
Compared to tables 3 and 4, the accuracy of behavior recognition is significantly higher than the position eigenvalues due to the influence of the height eigenvalues. To further demonstrate the best performance of the decision tree algorithm, the present invention compares naive bayes networks (NBN, naive Bayesiannetwork), random Forest (RF), random Forest (KNN, K-Nearest Neighbor). The behavior recognition error rates were compared with data sets D and 50 100 150 200, respectively, as shown in fig. 4 below:
according to fig. 4, the invention shows that the error rate of the decision tree algorithm DR is significantly lower than the other three algorithms, especially with the increase of the data set, the performance is better and the error rate is lower. When the algorithm is greatly influenced by the sensor, the KNN algorithm has the largest error rate due to the influence on the sensor; when d=100, the algorithm is similar to RF, and the KNN algorithm error rate is maximum; when d=100, the error rate of the RF algorithm is maximum, NBN is the second; the error rate of the RF algorithm is greatest when d=200.
According to the method for realizing behavior recognition based on the decision tree algorithm, firstly, the sensor is used for acquiring user data, then the extracted characteristic value is recognized, and finally, behavior classification experiment verification is realized by adopting the decision tree algorithm. Experimental results show that the algorithm has good daily behavior recognition performance under specific conditions.
The foregoing description of the preferred embodiments of the invention is not intended to limit the invention to the precise form disclosed, and any such modifications, equivalents, and alternatives falling within the spirit and scope of the invention are intended to be included within the scope of the invention.
Claims (1)
1. The accurate behavior recognition method based on the decision tree classification algorithm is characterized by comprising the following steps of:
s1, data collection: solving the label position by adopting a Chan algorithm based on TDOA;
s2, extracting features: selecting characteristic values and partitioning characteristic value intervals;
s3, behavior recognition: establishing a user behavior recognition model;
the step S1 specifically includes:
selecting 4 base stations; in the two-dimensional plane rectangular coordinate system, the coordinate of the ith base station is B uwb,i ＝[x i ,y i ] T (i=1, 2, …, 5) the coordinates of the tag are T uwb ＝[x 0 ,y 0 ] T A non-line of sight between the base station and the tag is R i ＝||B uwb,i -T uwb || 2 (i=1, 2, …, 5); taking a first base station as a common reference node to obtain a group of TDOA observation values delta t i,1 (i=2, 3,4, 5) indicating a signal arrival time difference between the i-th base station and the first base station;
in the case of this model of the present invention,is Deltat i,1 True value of n i,1 Measured by systematic error, NLOS error is n NLOS,i ；
Let the signal propagation velocity be c and calculate R i,1 Difference between the distances from the mark to the i-th base station and the first base station:
R i,1 ＝c·Δt i,1 (i＝2,3,4,5) (2)
establishing three hyperbolic equations R according to hyperbolic characteristics i,1 ＝R i -R 1 (i=2, 3,4, 5), regarding T uwb Can be established as shown in formula (3);
adopting a label positioning architecture of 4 base stations, taking one base station as a main part, and the rest 3 base stations; when a tester carrying a locating tag enters a test area, signals sent by the tag are received by one or more sensors; decoding signals from the sensors that transmit angle of arrival and timing information, and then transmitting these data to the master sensor; the main sensor collects all information sent by the base station to calculate the position information of the tag, so that data collection is realized; then the sensor transmits data every second through the exchanger and the server, the data adopts a UDP data packet format, the server receives the UDP data packet, and the X, Y coordinate information specific to the tag can be obtained;
the step S2 specifically includes:
the feature value selection includes: position division, head-shoulder waist-knee height processing and distance movement in unit time of head-shoulder waist and knee joint;
the location division includes:
in real life, the position of a user has a certain relation with the behavior activity of the user; spatial locations fall into three categories: first, the user can sit in a place where he lies for rest; the second type is a distance of 0.1 to 0.3 meters from the region, denoted Da, depending on the object; the remaining space is of the third class, denoted La;
the head, shoulder, waist and knee height treatment comprises head height, shoulder height, waist height and knee height;
z-axis data of the head, shoulder, waist and knee joints of a user represent the height of a user space and are directly read from tag coordinates;
the distance moves within unit time of the head, shoulder, waist and knee joints and comprises the distance from the head to the shoulder to the waistline and the knee;
direct calculation of the distance between the head, shoulder and waist of the user and the knee joint in unit time is difficult to achieve, mainly because the unit time is difficult to determine; the excessive numerical value influences the displacement calculation result, so that the user behavior cannot be accurately described, and the accuracy of the user behavior recognition error is reduced; too small a value increases the amount of computation due to delay overhead; through multiple experiments, the optimal unit time of weighing accuracy and operation delay is LS;
the partition characteristic value interval comprises:
after determining the above-mentioned feature values, classification boundaries must be determined to ensure similarity between data and intra-category, inter-category differences; combining experiments, and processing the mixture by adopting a layering classification method; classification focuses on how to determine the boundaries of each level; currently, determining boundary values is two algorithms: length equivalent method and distributed equivalent method;
set the range of eigenvalues phi= [ c ] min ,c max ]The method is divided into N grades, and grade labels are 1-N; from the value range phi= [ c min ,c max ]A range of sensor values r=c can be obtained max -c min The method comprises the steps of carrying out a first treatment on the surface of the To ensure that the length of each interval within the region is the same; then calculating to obtain the length of each interval as r=r/N; thus, a value range for each interval can be determined;
the step S3 specifically includes:
let D be the training tuples divided by category, then the entropy of D is expressed as:
wherein p is i Representing the probability that the ith class appears in the entire training tuple, can be estimated by dividing the number of elements belonging to that class by the total number of elements in the training tuple; the actual meaning of entropy represents the average amount of information required for tuple class labels in D;
let the training tuple D divided by the attribute A, the expected information for the D partition is:
the information delta is the difference between them:
gain(A)＝info(D)-info A (D) (6)
establishing a user behavior recognition model; the specific behavior recognition steps are as follows:
s31, classifying and collecting various behaviors according to a training data set, and dividing a training tuple into entropy info (D) of a training tuple;
s32, extracting position height characteristic values from data preprocessing, calculating characteristic value intervals and dividing the characteristic values;
s33, obtaining information gain (A) about the characteristic value of the partition before the prediction in the step 2;
s34, the difference between the information increment is caused by the expected information difference;
when the maximum value of the output is reached, memorizing the maximum gain as the maximum memory; the delta information is then the corresponding behavior information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911025926.5A CN110781961B (en) | 2019-10-25 | 2019-10-25 | Accurate behavior recognition method based on decision tree classification algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911025926.5A CN110781961B (en) | 2019-10-25 | 2019-10-25 | Accurate behavior recognition method based on decision tree classification algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110781961A CN110781961A (en) | 2020-02-11 |
| CN110781961B true CN110781961B (en) | 2024-02-23 |
Family
ID=69386768
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911025926.5A Active CN110781961B (en) | 2019-10-25 | 2019-10-25 | Accurate behavior recognition method based on decision tree classification algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110781961B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111339385A (en) * | 2020-02-26 | 2020-06-26 | 山东爱城市网信息技术有限公司 | CART-based public opinion type identification method and system, storage medium and electronic equipment |
| CN111417067B (en) * | 2020-03-13 | 2021-05-07 | 智慧足迹数据科技有限公司 | Method and device for positioning visited position of user |
| CN113112635B (en) * | 2021-04-12 | 2022-11-15 | 滁州博格韦尔电气有限公司 | Conventional inspection system for intelligent equipment |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108764282A (en) * | 2018-04-19 | 2018-11-06 | 中国科学院计算技术研究所 | A kind of Class increment Activity recognition method and system |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7714778B2 (en) * | 1997-08-20 | 2010-05-11 | Tracbeam Llc | Wireless location gateway and applications therefor |
| US9134398B2 (en) * | 1996-09-09 | 2015-09-15 | Tracbeam Llc | Wireless location using network centric location estimators |
| US20030146871A1 (en) * | 1998-11-24 | 2003-08-07 | Tracbeam Llc | Wireless location using signal direction and time difference of arrival |
-
2019
- 2019-10-25 CN CN201911025926.5A patent/CN110781961B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108764282A (en) * | 2018-04-19 | 2018-11-06 | 中国科学院计算技术研究所 | A kind of Class increment Activity recognition method and system |
Non-Patent Citations (2)
| Title |
|---|
| 张聪聪 ; 宋承祥 ; 丁艳辉 ; .群体仿真中基于决策树的路径自动评价研究.计算机工程与设计.2015,(02),全文. * |
| 李孝伟 ; 陈福才 ; 李邵梅 ; .基于分类规则的C4.5决策树改进算法.计算机工程与设计.2013,(12),全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110781961A (en) | 2020-02-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110781961B (en) | Accurate behavior recognition method based on decision tree classification algorithm | |
| US10884112B2 (en) | Fingerprint positioning method and system in smart classroom | |
| Ge et al. | Optimization WIFI indoor positioning KNN algorithm location-based fingerprint | |
| Ngamakeur et al. | A survey on device-free indoor localization and tracking in the multi-resident environment | |
| CN103905992B (en) | Indoor positioning method based on wireless sensor networks of fingerprint data | |
| CN105717483B (en) | A kind of location determining method and device based on multi-source positioning method | |
| CN103561463A (en) | RBF neural network indoor positioning method based on sample clustering | |
| CN103281779B (en) | Based on the radio frequency tomography method base of Background learning | |
| CN106353726A (en) | Twice-weighted mass center determining method and system for indoor positioning | |
| Zhang et al. | Indoor 2.5 D positioning of WiFi based on SVM | |
| CN110213710A (en) | A kind of high-performance indoor orientation method, indoor locating system based on random forest | |
| Cheng et al. | Activity recognition and localization based on UWB indoor positioning system and machine learning | |
| CN109541537B (en) | Universal indoor positioning method based on ranging | |
| CN112533136A (en) | WLAN fingerprint positioning method based on deep learning | |
| Guo et al. | TWCC: A robust through-the-wall crowd counting system using ambient WiFi signals | |
| Sridharan et al. | Inferring micro-activities using wearable sensing for ADL recognition of home-care patients | |
| Martín et al. | Affinity propagation clustering for older adults daily routine estimation | |
| Al-Husseiny et al. | Unsupervised learning of signal strength models for device-free localization | |
| CN105960011B (en) | Indoor objects localization method based on Sensor Network and bayes method | |
| Jain et al. | Performance analysis of received signal strength fingerprinting based distributed location estimation system for indoor wlan | |
| Zhao et al. | A nlos detection method based on machine learning in uwb indoor location system | |
| Salsabila et al. | The implementation of optimal k-means clustering for indoor moving object localization | |
| Lv et al. | Research on Indoor Mobile Node Localization Based on Wireless Sensor Networks. | |
| Münch et al. | Towards a device-free passive presence detection system with Bluetooth Low Energy beacons. | |
| CN112261718B (en) | Indoor positioning method |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |