CN114758408A - Multi-classifier-based confusion behavior conflict management system and method - Google Patents
Multi-classifier-based confusion behavior conflict management system and method Download PDFInfo
- Publication number
- CN114758408A CN114758408A CN202011589114.6A CN202011589114A CN114758408A CN 114758408 A CN114758408 A CN 114758408A CN 202011589114 A CN202011589114 A CN 202011589114A CN 114758408 A CN114758408 A CN 114758408A
- Authority
- CN
- China
- Prior art keywords
- behavior
- confusion
- conflict
- classifiers
- information
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 24
- 230000006399 behavior Effects 0.000 claims abstract description 256
- 238000001514 detection method Methods 0.000 claims abstract description 34
- 230000008030 elimination Effects 0.000 claims abstract description 22
- 238000003379 elimination reaction Methods 0.000 claims abstract description 22
- 238000007781 pre-processing Methods 0.000 claims abstract description 12
- 238000006243 chemical reaction Methods 0.000 claims abstract description 7
- 230000005540 biological transmission Effects 0.000 claims description 33
- 238000007726 management method Methods 0.000 claims description 29
- 230000004927 fusion Effects 0.000 claims description 23
- 230000001133 acceleration Effects 0.000 claims description 15
- 238000013507 mapping Methods 0.000 claims description 15
- 238000010606 normalization Methods 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 11
- 230000010365 information processing Effects 0.000 claims description 11
- 239000011159 matrix material Substances 0.000 claims description 11
- 239000013598 vector Substances 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 7
- 238000013145 classification model Methods 0.000 claims description 6
- 230000006698 induction Effects 0.000 claims description 6
- 230000009466 transformation Effects 0.000 claims description 6
- 239000003550 marker Substances 0.000 claims description 5
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 claims description 4
- 239000008280 blood Substances 0.000 claims description 4
- 210000004369 blood Anatomy 0.000 claims description 4
- 229910052760 oxygen Inorganic materials 0.000 claims description 4
- 239000001301 oxygen Substances 0.000 claims description 4
- 230000001960 triggered effect Effects 0.000 claims description 3
- 238000009826 distribution Methods 0.000 claims description 2
- 230000006870 function Effects 0.000 description 11
- 238000010586 diagram Methods 0.000 description 9
- 238000004422 calculation algorithm Methods 0.000 description 7
- 238000004590 computer program Methods 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 6
- 238000013459 approach Methods 0.000 description 4
- 238000003860 storage Methods 0.000 description 4
- QVFWZNCVPCJQOP-UHFFFAOYSA-N chloralodol Chemical compound CC(O)(C)CC(C)OC(O)C(Cl)(Cl)Cl QVFWZNCVPCJQOP-UHFFFAOYSA-N 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000003909 pattern recognition Methods 0.000 description 2
- 230000000717 retained effect Effects 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000003542 behavioural effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000000276 sedentary effect Effects 0.000 description 1
Images
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
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention provides a confusion behavior conflict management system and method based on multiple classifiers, which are used for collecting physiological information and motion information of a measured object; transmitting and storing physiological information and motion information; preprocessing physiological information and motion information to obtain input data of a plurality of classifiers, and inputting the preprocessed data into the plurality of classifiers for behavior recognition; obtaining classification results of a plurality of classifiers, determining whether confusion behaviors exist among the classification results and the types of the confusion behaviors, if so, triggering a conflict detection and elimination task, wherein the conflict detection task carries out marking conversion by obtaining the probability of real labels and behavior information to determine whether conflicts exist, and the elimination task eliminates the conflicts among different classifiers; the recognition accuracy of the confusion behavior after the conflict is eliminated is calculated. The invention can improve the accuracy of behavior recognition.
Description
Technical Field
The invention belongs to the technical field of pattern recognition and data fusion, and particularly relates to a multi-classifier-based confusion behavior conflict management system and method.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
A Human Activity Recognition (HAR) system is a system that obtains various kinds of behavior information of a Human body and implements behavior Recognition by using a reasonable model. During the past decade, the concept and technology of the internet of things has been rapidly developed by integrating traditional networks, wearable device sensors, and networked objects. Meanwhile, the human behavior recognition technology is advanced to a new height due to the gradual maturity of mainstream technologies such as big data and cloud computing. Nowadays, human behavior recognition technology has been widely applied in smart home, electronic health, intelligent medical diagnosis, and elderly care. At present, algorithms such as machine learning and deep learning are mainly adopted in the behavior recognition model.
The development of human behavior recognition technology has made great progress so far, and has made certain breakthrough in both the feature acquisition layer and the complex behavior recognition layer. However, the behavior recognition algorithm is not yet mature, and at present, no algorithm suitable for all behavior classifications exists. On one hand, a single classifier cannot achieve a high recognition rate with low complexity for human behaviors, and therefore, the single classifier often performs a wrong recognition on some features and similar behaviors, for example: recognizing 'standing' as 'sitting still', recognizing 'going upstairs' as 'going downstairs' and the like, wherein the confusion behaviors can cause the reduction of the recognition accuracy; on the other hand, there is also a large difference between classifiers of different kinds. Thus, experts in the field of pattern recognition gradually have conducted deep research into multi-classifier fusion algorithms. The comprehensive use of the single classifier enables the fused multiple classifiers to have prominent advantages and overcome defects, and classification precision and operation speed are obviously improved.
At present, a human behavior recognition system lacks an effective judgment mode for a confusion behavior, although a human behavior recognition technology combines a multi-classifier fusion model and an algorithm, the recognition accuracy of the confusion behavior can be directly influenced by conflicts generated by fusion of multiple classifiers aiming at the confusion behavior. Therefore, how to eliminate the conflict among multiple classifications through a reasonable fusion algorithm is a challenge to be solved for improving the accuracy of identifying the confusion behavior.
Disclosure of Invention
The invention provides a confusion behavior conflict management system and method based on multiple classifiers in order to solve the problems, and the system and method can solve the problem of low recognition accuracy caused by conflict existing among the multiple classifiers aiming at confusion behaviors and improve the accuracy of behavior recognition.
According to some embodiments, the invention adopts the following technical scheme:
a multi-classifier based confusion behavior conflict management system, comprising:
the behavior information acquisition module comprises a plurality of sensors distributed at different positions of the measured object and is used for acquiring physiological information and motion information of the measured object in real time;
the behavior information transmission module is configured to transmit the information acquired by the behavior information acquisition module to the behavior information processing module;
the behavior information processing module is configured to preprocess the physiological information and the motion information to obtain input data of a plurality of classifiers and input the preprocessed data into the plurality of classifiers for behavior recognition;
the confusion behavior judging module is configured to obtain the classification results of the plurality of classifiers, determine whether confusion behaviors exist among the classification results and determine the type of the confusion behaviors, and if so, send a signal to the conflict management module;
and the conflict management module is configured to trigger a conflict detection and elimination task according to the received signal, wherein the conflict detection task performs marking conversion by acquiring the probability of the real label and the behavior information, determines whether a conflict exists, eliminates the conflict among different classifiers by the elimination task, and calculates the identification accuracy of the confusion behavior after the conflict is eliminated.
As an alternative embodiment, the physiological information includes heart rate, blood oxygen and skin resistance value, and the motion information includes three-axis acceleration, three-axis angular velocity and three-axis magnetic induction information.
The three-axis acceleration comprises X-axis acceleration, Y-axis acceleration and Z-axis acceleration; the three-axis angular velocity comprises an X-axis angular velocity, a Y-axis angular velocity and a Z-axis angular velocity; the three-axis magnetic induction information includes vectors in the three axes X, Y, Z of the resultant magnetic field generated by the earth's magnetic field and the circuit board device in the location.
As an alternative embodiment, the behavior information transmission module transmits the physiological information and the motion information of the measured object to the cloud server, and the behavior information processing module acquires the physiological information and the behavior information from the cloud server.
As an alternative implementation manner, the behavior information transmission module includes a behavior information transmission mode selection unit and a behavior information wireless transmission unit, which are connected in sequence, wherein:
the behavior information transmission mode selection unit is used for selecting a proper wireless transmission mode for transmission;
the behavior information wireless transmission unit is used for transmitting the behavior information according to the selected wireless transmission mode.
As an alternative embodiment, the behavior information processing module includes a behavior information preprocessing unit and a multi-classifier classifying unit, wherein:
the behavior information preprocessing unit carries out preprocessing operation according to the physiological information and the motion information to obtain input data matched with the multiple classifiers;
and the multi-classifier classification unit is used for classifying and identifying the preprocessed data in parallel and obtaining the identification result of each classifier.
The confusion behavior determination module comprises a threshold triggering unit and a behavior information output unit, wherein:
the threshold triggering unit is configured to calculate the probability that each classifier is in descending order to generate false recognition, and if the value exceeds a preset threshold, the behavior information output unit is triggered by considering that more serious confusion exists among different behaviors;
the behavior information output unit is configured to output the specifically confused behavior types and the probabilities that different classifiers are the predetermined types of behaviors to be identified.
By way of further limitation, the preset threshold is adaptively updated according to the recognition results of different classifier combinations and the confusion matrix.
As an alternative embodiment, the conflict management module includes a conflict detection unit and a conflict elimination unit connected in sequence, where:
the conflict detection unit is used for detecting whether conflicts exist among the multiple classifiers aiming at the confusion behavior.
The conflict elimination unit is used for eliminating conflicts among the multiple classifiers aiming at the confusion behavior and obtaining a final fusion result.
As a further limitation, the collision detection unit includes a true tag and behavior information probability acquisition subunit, a labeled transformation subunit, and a multi-classifier collision detection subunit, which are connected in sequence, wherein:
the real label and behavior information probability obtaining subunit is used for obtaining Softmax layer output probabilities of different classifiers;
the marked transformation module is used for determining the output probability of the Softmax layer to be obtained according to the row index value of the confusion behavior in the real label, and converting the judgment results of the behavior types output by the three classification models into different marked types to be used as an identification frame of a subsequent D-S evidence theory;
the multi-classifier collision detection subunit is configured to detect whether a collision exists between the multi-classifiers for the confusion behavior.
As a further limitation, the conflict elimination unit comprises an entropy calculation subunit, an entropy and reliability mapping subunit, a reliability normalization subunit and a D-S evidence theory fusion judgment subunit which are connected in sequence, wherein the behavior information probability acquisition subunit is connected with the entropy calculation subunit, and the multi-classifier conflict detection subunit is connected with the D-S evidence theory fusion judgment subunit;
the entropy calculation subunit is used for calculating entropy values of different classifiers for the confusion behavior probability;
the entropy value and reliability mapping subunit obtains the reliability of the focal elements of different classifiers under an identification frame through a mapping relation established between the entropy value and the uncertain focal elements;
the credibility normalization subunit is used for normalization processing of the credibility of the focal elements under different classifiers, so that basic conditions of the D-S evidence theory are met;
and the D-S evidence theory fusion judgment subunit is used for eliminating conflicts among different classifiers according to the evidence combination rule and acquiring a final fusion judgment result.
A confusion behavior conflict management method based on multiple classifiers comprises the following steps:
collecting physiological information and motion information of a measured object;
transmitting and storing physiological information and motion information;
preprocessing physiological information and motion information to obtain input data of a plurality of classifiers, and inputting the preprocessed data into the plurality of classifiers for behavior recognition;
obtaining classification results of a plurality of classifiers, determining whether confusion behaviors exist among the classification results and the types of the confusion behaviors, if so, triggering a conflict detection and elimination task, wherein the conflict detection task carries out marking conversion by obtaining the probability of real labels and behavior information to determine whether conflicts exist, and the elimination task eliminates the conflicts among different classifiers;
and calculating the recognition accuracy of the confusion behavior after the conflict is eliminated.
As an alternative embodiment, the specific process of determining whether the confusion behavior exists between the classification results and the type of the confusion behavior includes:
and analyzing the confusion matrix obtained by different classifiers, judging whether confusion behaviors exist or not according to a preset threshold value and outputting the type of the confusion behaviors, and if the confusion behaviors exceed the preset threshold value, judging that the confusion behaviors exist.
As an alternative embodiment, the conflict detection and elimination task process includes:
acquiring real labels of behaviors to be identified, determining column indexes of the confused behaviors according to the real labels, and acquiring probability vectors of various behaviors output by different classifiers according to the indexes;
calculating a column index corresponding to the maximum probability value according to the probability vectors of different types of behaviors in different classifiers, and marking correspondingly;
determining whether conflict detection exists between different classifiers according to the marker sequences of the different classifiers, and if not, selecting the marker sequence of any classifier as an output result;
and if the result exists, acquiring entropy values of the uncertain focal elements by different classifiers, selecting a proper mapping relation according to the acquired entropy values to obtain initial reliability distribution of the uncertain focal elements, carrying out normalization processing on the initial reliability of each focal element, carrying out fusion judgment on the normalized result and outputting a final judgment result.
Compared with the prior art, the invention has the following beneficial effects:
the behavior information acquisition mode can be actively selected so as to adapt to the acquisition requirements of different scenes;
the method can judge the occurrence and specific type of the confusion behavior in real time according to the classification results of different classifiers and the confusion matrix, and detect and eliminate the conflict among the multiple classifiers in real time according to the D-S evidence theory, thereby improving the identification accuracy of the confusion behavior.
The single classifier not only pays a large cost in the aspects of algorithm time consumption and space complexity for improving the identification accuracy rate aiming at the confusion behavior, but also is not beneficial to the improvement of the overall performance of the system. According to the method, the conflict existing among different classifiers aiming at the confusion behavior is eliminated according to the D-S evidence theory, the recognition accuracy of the confusion behavior is improved, and the energy consumption of the system is reduced.
In order to make the aforementioned and other objects, features and advantages of the present invention comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.
Fig. 1 is a diagram of module composition and connection relationship of a multi-classifier based confusion behavior conflict management system according to the present invention.
FIG. 2 is a structural framework diagram of a multi-classifier based confusion behavior conflict management system according to the present invention.
FIG. 3 is a flowchart illustrating the operation of the confusion behavior conflict management system based on multiple classifiers according to the present invention.
FIG. 4 is a confusion matrix diagram of the multiple classifiers in processing the WISDM data set according to the confusion behavior conflict management system based on the multiple classifiers.
The specific implementation mode is as follows:
the invention is further described with reference to the following figures and examples.
It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
Example 1
A multi-classifier based confusion behavior conflict management system, as shown in fig. 1, comprising: the system comprises a behavior information acquisition module, a behavior information transmission module, a behavior information processing module, a confusion behavior judgment module and a conflict management module.
The behavior information acquisition module, the behavior information transmission module, the behavior information processing module, the confusion behavior judgment module and the conflict management module are sequentially connected.
The behavior information acquisition module is used for acquiring physiological information and behavior information of a human body in real time, wherein the physiological information comprises heart rate, blood oxygen and skin resistance value. The motion information includes three-axis acceleration, three-axis angular velocity, and three-axis magnetic induction information. The three-axis acceleration comprises X-axis acceleration, Y-axis acceleration and Z-axis acceleration; the three-axis angular velocity comprises an X-axis angular velocity, a Y-axis angular velocity and a Z-axis angular velocity; the three-axis magnetic induction information includes vectors in the three axes X, Y, Z of the resultant magnetic field generated by the earth's magnetic field and the circuit board device in the location. The behavior information transmission module is used for transmitting the collected human physiological information and the collected motion information to the cloud server. The behavior information processing module is used for acquiring human physiological information and behavior information of the cloud server, preprocessing the information to obtain input data of the plurality of classifiers, and inputting the preprocessed data into the plurality of classifiers for behavior recognition. The confusion behavior judging module is used for judging whether confusion behaviors exist and the types of the confusion behaviors in real time according to the classification result of the multi-classifier. And the conflict management module is used for triggering conflict detection and elimination and obtaining the identification accuracy of the confusion behavior after the conflict is eliminated.
As shown in fig. 2, the behavior information collecting module includes an information type selecting unit, a physiological information sensing unit, and a motion information sensing unit.
The information type selection unit is respectively connected with the physiological information sensing unit and the motion information sensing unit. The information type selection unit is used for triggering the physiological information sensing unit or the motion information sensing unit.
The physiological information sensing unit is used for acquiring physiological information of a human body, such as heart rate, blood oxygen, skin resistance value and the like.
The motion information sensing unit is used for acquiring nine-axis motion information of the human body, wherein the nine-axis motion information comprises three-axis acceleration, three-axis angular velocity and three-axis magnetic induction information.
According to the invention, the behavior information transmission module preferably comprises a behavior information transmission mode selection unit and a behavior information wireless transmission unit which are sequentially connected.
The behavior information transmission mode selection unit is used for selecting different types of transmission modes such as LoRa, WiFi, 4G/5G, and the like, but is not limited to the wireless transmission mode.
The behavior information wireless transmission unit is used for transmitting the behavior information according to the selected wireless transmission mode.
According to the invention, the behavior information processing module preferably comprises a behavior information preprocessing unit and a multi-classifier classifying unit which are connected in sequence.
The behavior information preprocessing unit carries out preprocessing operations such as sliding window segmentation and the like according to the human physiological information and the motion information acquired from the cloud server side so as to obtain input data of the multiple classifiers.
And the multi-classifier classification unit is used for classifying and identifying the preprocessed data in parallel and obtaining the identification result of each classifier.
According to the present invention, preferably, the confusion behavior determination module includes a threshold triggering unit and a behavior information output unit which are connected in sequence.
The threshold triggering unit is used for calculating the probability that each classifier is in the descending state for the occurrence of false recognition, and if the value exceeds a preset threshold value of the system, the system considers that more serious confusion exists among different behaviors and triggers the behavior information output unit. For example, the system selects three different classifiers Conv2D, google lene and ResNet to perform behavior recognition on input data, the three classifiers output three different confusion matrices, and the system performs analysis according to the three different confusion matrices. Taking the confusion matrix output by the Conv2D as an example, the recognition accuracy rates of six behaviors of "going downstairs", "jogging", "sitting still", "standing", "going upstairs" and "walking" are respectively 85.9%, 96.0%, 97.8%, 98.0%, 87.7% and 98.7%, the recognition accuracy rates of "going downstairs" and "going upstairs" are low, and the recognition accuracy rates of four behaviors of "jogging", "sitting still", "standing" and "walking" are high, although there may be a certain degree of confusion among the four behaviors, the confusion degree is very low compared with the two behaviors of "going downstairs" and "going upstairs". Meanwhile, the probability of identifying "going downstairs" as "going upstairs" is 12.6%, and the probability of identifying "going upstairs" as "going downstairs" is 7.6%. Assuming that the threshold value preset by the system is 15%, the sum of the recognition error probabilities of the two behaviors of going upstairs and downstairs is 20.2%, and exceeds the threshold value preset by the system, so that the threshold value triggering unit is triggered, and the system judges that the behaviors of going upstairs and downstairs are confused in real time. The threshold is not fixed, since different classifier combinations will output different confusion matrices. For each preprocessed data, the system adaptively updates the threshold according to the recognition results of different classifier combinations and the confusion matrix.
The behavior information output unit is used for outputting specific confused behavior types and probabilities of different classifiers for various behaviors through the Softmax function.
According to the invention, the conflict management module comprises a conflict detection unit and a conflict elimination unit which are connected in sequence.
The conflict detection unit is used for detecting whether conflicts exist among the multiple classifiers aiming at the confusion behavior.
The conflict elimination unit is used for eliminating the conflict among the multiple classifiers aiming at the confusion action and obtaining the final fusion result.
According to the optimization of the invention, the conflict detection unit comprises a real label and behavior information probability acquisition subunit, a marked transformation subunit and a multi-classifier conflict detection subunit which are connected in sequence.
And the real label and behavior information probability acquiring subunit is used for acquiring Softmax layer output probabilities of different classifiers.
And the marked transformation unit is used for determining the output probability of the Softmax layer to be acquired according to the row index value of the confusion behavior in the real label, and converting the judgment results of the behavior types output by the three classification models into three marked types of '0', '1' and '2', and using the three marked types as the identification framework of the subsequent D-S evidence theory. Where "0" means "going upstairs", "1" means "going downstairs", and "2" means other types of behavior that are identified as neither "going upstairs" nor "going downstairs".
The multi-classifier collision detection subunit is configured to detect whether a collision exists between the multi-classifiers for the confusion behavior.
According to the optimization of the invention, the conflict elimination unit comprises an entropy calculation subunit, an entropy and reliability mapping subunit, a reliability normalization subunit and a D-S evidence theory fusion judgment subunit which are connected in sequence. Meanwhile, the behavior information probability obtaining subunit is connected with the entropy value calculating subunit.
The entropy calculation subunit is used for calculating entropy values of different classifiers for the confusion behavior probability.
And the entropy value and reliability mapping subunit obtains the reliability of the focal elements of different classifiers under the identification frame through the mapping relation established between the entropy value and the uncertain focal elements.
And the credibility normalization subunit is used for normalization processing of the credibility of the focal elements under different classifiers, so that the basic conditions of the D-S evidence theory are met.
And the D-S evidence theory fusion judgment subunit is used for eliminating conflicts among different classifiers according to the evidence combination rule and acquiring a final fusion judgment result.
Example 2
In this embodiment, a working method of the confusion behavior conflict management system based on multiple classifiers in embodiment 1 is provided, as shown in fig. 3, the system actively selects a motion information sensing unit to obtain a three-axis acceleration of a human body, and performs human behavior recognition on preprocessed data in parallel through three classifiers, namely Conv2D, google lenet and ResNet.
Step S01: behavior information acquisition mode selection
The behavior information acquisition mode set by the system is a motion information acquisition mode. According to the requirement of a behavior information acquisition mode required by behavior identification, an accelerometer, an angular velocity meter and a magnetometer are selected as sensors of motion information, and the motion information is acquired.
Step S02: behavior information wireless transmission mode selection
The behavior information wireless transmission modes set by the system include LoRa transmission, WiFi transmission, 4G/5G transmission and the like, but are not limited to the above wireless transmission modes. Assuming that the wireless transmission mode of the behavior information set by the system is LoRa transmission, the cloud server stores the transmitted behavior information.
Step S03: wireless transmission of behavioral information
And the system transmits the behavior information according to the selected wireless transmission mode.
Step S04: cloud server storage
And the cloud server stores the transmitted behavior information in real time.
Step S05: behavior information acquisition
According to the system requirements, the motion information can be obtained at the cloud server side for behavior recognition or analyzed by obtaining the motion information.
Step S06: behavior information preprocessing
And carrying out preprocessing operations such as sliding window segmentation on the human physiological information and the motion information acquired from the cloud server side. After being dividedEach data block represents a complete activity. For example, behavior recognition is based only on tri-axial acceleration, in which case the data can be expressed as X ∈ Rn×3Where n represents the total amount of X-axis (Y-axis or Z-axis) data, and 3 represents triaxial data using acceleration. The system can preset the size of a sliding window and the sliding mode of the window, if the sampling frequency of the sensor is f, the size selectable range of the sliding window is generally 0.5f-10f, and the sliding mode of the window is defaulted to be 50% of the window overlapping rate. Under this parameter setting, the data can be represented as Xi∈Rnwindow×3，i∈[1,N]Where N represents the total number of divided data blocks, NwindowIndicating the size of the sliding window.
Step S07: multi-classifier classification and identification
And inputting the preprocessed data into different classifiers, such as a Conv2D model, a GoogleNet model and a ResNet model, in parallel, so as to obtain classification results and confusion matrixes under different classifiers.
Step S08: real-time determination of confusion behavior
The system analyzes the confusion matrix obtained by different classifiers, judges whether confusion behaviors exist or not according to a preset threshold value and outputs the types of the confusion behaviors. For example, the recognition accuracy of the Conv2D model for six behaviors of "going downstairs", "jogging", "sitting still", "standing", "going upstairs" and "walking" is 85.9%, 96.0%, 97.8%, 98.0%, 87.7% and 98.7%, respectively, and confusion is highly likely to occur when the recognition accuracy for "going downstairs" and "going upstairs" is low. Meanwhile, the probability of identifying "going downstairs" as "going upstairs" is 12.6%, and the probability of identifying "going upstairs" as "going downstairs" is 7.6%. Assuming that the threshold value preset by the system is 15%, the sum of the probability of the recognition errors of the two behaviors is 20.2%, and the threshold value preset by the system is exceeded, so that the system judges that the behaviors of going upstairs and going downstairs are confused in real time.
Step S09: true tag and behavior information probability acquisition
The system first obtains the real label and determines the column index where the confusion action occurs according to the real label. For example, there are six behaviors to be identified for the WISDM dataset, which are "downstairs," "jogging," "sedentary," "standing," "upstairs," and "walking," respectively. The real label is a matrix with dimensions of m × 6, where m represents the number of a complete behavior cycle, and 6 represents six types of behaviors. And outputting the final recognition result of the classification model by each row of the matrix in a one-hot coding mode. For example, [ 100000 ] is a line derived from a real tag, which represents that the recognized behavior is "downstairs". According to step S08, the act of confusion is "downstairs" and "upstairs", the column indices of which in the real label are 1 and 5, respectively, and thus "downstairs" is encoded as 1 and "upstairs" is encoded as 0. Meanwhile, the system carries out XOR operation on the codes of the two columns, and if the XOR result is 1, the corresponding row index is reserved; otherwise, it is discarded. For example, "downstairs" is coded as 1, and "upstairs" is coded as 0, and the result of xor of the two is 1, at this time, the system retains the corresponding row index, and obtains the probability vectors of the six types of behaviors output by the Softmax layers of different classifiers according to the retained row index. For example, the retained row index is 15, the output probability of the Conv2D model corresponding thereto passing through the Softmax function for six types of behaviors is [ 0.97490.00010.00000.00000.02490.0001 ], the output probability of the google net model passing through the Softmax function for six types of behaviors is [ 0.37670.00000.00000.00000.62330.0000 ], and the output probability of the ResNet model passing through the Softmax function for six types of behaviors is [ 0.92450.00010.00000.00000.07530.0000 ]. Thus, the system completes the acquisition of the behavior information probability.
Step S10: marker conversion
The system calculates a column index corresponding to the maximum probability value of each row according to probability vectors of different types of behaviors in different classifiers, if the column index is 5, the column index is marked as 0, if the column index is 1, the column index is marked as 1, and if the column index is 2, the frame Θ is identified as {0,1,2 }. For example, the Conv2D model has an output probability of [ 0.38500.00430.00010.00010.58100.0295 ] for six types of behaviors through the Softmax function, the GoogleLeNet model has an output probability of [ 0.14420.06200.00010.00010.37560.4180 ] for six types of behaviors through the Softmax function, and the ResNet model has an output probability of [ 0.67880.01310.00010.00000.26150.0465 ] for six types of behaviors through the Softmax function. The Conv2D model has a column index of 5 for the maximum of the probability vectors for the six classes of behavior, and is therefore labeled "0"; the GoogleLeNet model has a column index of 6 corresponding to the maximum value of the probability vectors of the six classes of behaviors, and therefore is marked as "2"; the ResNet model has a column index of 1 for the maximum value of the six classes of behavior probability vectors, and is therefore labeled "1". Thus, each classifier was subjected to label conversion to obtain a labeled sequence containing "0", "1", and "2".
Step S11: multi-classifier collision detection
The system carries out conflict detection on the mark sequences of different classifiers, if the sum of each column is 3 or 0, the situation that no conflict exists between the different classifiers aiming at the confusion behavior is judged, and if not, the situation that the conflict exists between the different classifiers aiming at the confusion behavior is judged.
Step S12: entropy calculation
Due to 2ΘThe information source entropy in the information theory can quantify the degree of uncertainty of the information source, wherein the entropy of the information source comprises {0,1,2, {0,1}, {0,2}, {1,2}, {1,0,2}, phi }, focal elements {0,1}, {0,2}, {1,2}, and {0,1,2 }. Thus, in D-S evidence theory, the entropy values of the computed focal elements 0,1, 0,2, 1,2, 0,1,2 may particularly represent the degree of "uncertainty". For example, the probability of the focal element {1} is p ({1}), the probability of the focal element {0} is p ({0}), and the probability of the focal element {2} is 1-p ({1}) -p ({0}), so the entropy of the focal element {0,1,2} is calculated as follows:
H({0,1,2})＝-p({0})·log3[p({0})]-p({1})·log3[p({1})]-p({2})·log3[p({2})](1)
similarly, the entropy values of other focal elements can be calculated according to the above formula. Thus, the system may obtain entropy values for different classifiers for uncertain focal elements.
Step S13: entropy and confidence mapping
The system selects a proper mapping relation according to the obtained entropy value to obtain the initial reliability allocation (BPA) of the uncertain focus elements. For example, an exponential type mapping relationship
f(H)＝H·e1-H(H∈[0,1]) (2)
Satisfies the range between [0,1], when the entropy value H approaches 0, f (H) approaches 0, when the entropy value H approaches 1, f (H) approaches 1 and when the derivative of f (H) is 0. For example, the BPA calculation for focal elements {0,1,2} is as follows:
m({0,1,2})＝H({0,1,2})·e1-H({0,1,2}) (3)
therefore, through the mapping relation, the system can acquire the BPA of uncertain focal elements of different classifiers.
Step S14: confidence normalization
To satisfy the basic initial conditions of D-S evidence theory, the system needs to normalize BPA for each focal bin, for example, BPA normalized by focal bin {0,1,2} is as follows:
therefore, the system can obtain the BPA normalized by each focal element of different classifiers according to the process.
Therefore, the system can obtain the BPA normalized by each focal element of different classifiers according to the process.
Step S15: D-S evidence theory fusion decision
If no conflict exists among the multiple classifiers, the system selects the mark sequence of any one classifier as an output result; and if conflicts exist among the multiple classifiers, the system performs fusion judgment according to the acquired BPA after each focal element of the different classifiers is normalized and outputs a final judgment result.
Step S16: calculation of recognition accuracy rate for eliminating confusion behavior after multi-classifier conflict
The system actively calculates the recognition accuracy rate of the confusion behavior after eliminating the conflict among the multiple classifiers aiming at the confusion behavior. For example, the recognition accuracy of the Conv2D model for the "downstairs" behavior is 85.90%, and the recognition accuracy for the "upstairs" behavior is 87.70%; the identification accuracy of the GoogLeNet model for the behavior of going downstairs is 87.35%, and the identification accuracy of the GoogLeNet model for the behavior of going upstairs is 87.02%; the recognition accuracy of the ResNet model for the behavior of going downstairs is 80.37%, and the recognition accuracy for the behavior of going upstairs is 85.10%. After the D-S evidence theory eliminates the conflict between different classification models, the identification accuracy rate of the behavior of going downstairs is 88.41%, the identification accuracy rate of the behavior of going upstairs is 92.89%, and the identification accuracy rates of the two types of confused behaviors are obviously improved compared with a single classification model.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Although the embodiments of the present invention have been described with reference to the accompanying drawings, it is not intended to limit the scope of the invention, and it should be understood by those skilled in the art that various modifications and variations can be made without inventive changes in the technical solutions of the present invention.
Claims (10)
1. A confusion behavior conflict management system based on multiple classifiers is characterized in that: the method comprises the following steps:
the behavior information acquisition module comprises a plurality of sensors distributed at different positions of the measured object and is used for acquiring physiological information and motion information of the measured object in real time;
the behavior information transmission module is configured to transmit the information acquired by the behavior information acquisition module to the behavior information processing module;
the behavior information processing module is configured to preprocess the physiological information and the motion information to obtain input data of a plurality of classifiers and input the preprocessed data into the plurality of classifiers for behavior recognition;
the confusion behavior judging module is configured to obtain the classification results of the plurality of classifiers, determine whether confusion behaviors exist among the classification results and determine the type of the confusion behaviors, and if so, send a signal to the conflict management module;
and the conflict management module is configured to trigger a conflict detection and elimination task according to the received signal, wherein the conflict detection task performs marking conversion by acquiring the probability of the real label and the behavior information, determines whether a conflict exists, eliminates the conflict among different classifiers by the elimination task, and calculates the identification accuracy of the confusion behavior after the conflict is eliminated.
2. The multi-classifier based confusion behavior conflict management system of claim 1, wherein: the physiological information comprises heart rate, blood oxygen and skin resistance values, and the motion information comprises three-axis acceleration, three-axis angular velocity and three-axis magnetic induction information.
3. The multi-classifier based confusion behavior conflict management system of claim 1, wherein: the behavior information transmission module transmits the physiological information and the motion information of the tested object to the cloud server, and the behavior information processing module acquires the physiological information and the behavior information from the cloud server.
4. The multi-classifier based confusion behavior conflict management system of claim 1, wherein:
the confusion behavior determination module comprises a threshold triggering unit and a behavior information output unit, wherein:
the threshold triggering unit is configured to calculate the probability that each classifier is in descending order to generate false recognition, and if the value exceeds a preset threshold, the behavior information output unit is triggered by considering that more serious confusion exists among different behaviors;
the behavior information output unit is configured to output specific confused behavior types and probabilities that different classifiers are various predetermined behaviors to be recognized;
or further, the preset threshold value is adaptively updated according to the recognition results of different classifier combinations and the confusion matrix.
5. The multi-classifier based confusion behavior conflict management system of claim 1, wherein: the conflict management module comprises a conflict detection unit and a conflict elimination unit which are connected in sequence, wherein:
the conflict detection unit is used for detecting whether conflicts exist among the multiple classifiers aiming at the confusion behavior.
The conflict elimination unit is used for eliminating conflicts among the multiple classifiers aiming at the confusion behavior and obtaining a final fusion result.
6. The multi-classifier based confusion behavior conflict management system of claim 1, wherein: the conflict detection unit comprises a real label and behavior information probability acquisition subunit, a marked transformation subunit and a multi-classifier conflict detection subunit which are sequentially connected, wherein:
the real label and behavior information probability obtaining subunit is used for obtaining Softmax layer output probabilities of different classifiers;
the marked transformation unit is used for determining the output probability of the Softmax layer to be obtained according to the row index value of the confusion behavior in the real label, and converting the judgment results of the behavior types output by the three classification models into different marked types to be used as an identification frame of a subsequent D-S evidence theory;
the multi-classifier conflict detection subunit is configured to detect whether a conflict exists between the multi-classifiers for the confusion behavior.
7. The multi-classifier based confusion behavior conflict management system of claim 1, wherein: the conflict elimination unit comprises an entropy calculation subunit, an entropy and reliability mapping subunit, a reliability normalization subunit and a D-S evidence theory fusion judgment subunit which are sequentially connected, the behavior information probability acquisition subunit is connected with the entropy calculation subunit, and the multi-classifier conflict detection subunit is connected with the D-S evidence theory fusion judgment subunit;
the entropy calculation subunit is used for calculating entropy values of different classifiers for the confusion behavior probability;
the entropy value and reliability mapping subunit obtains the reliability of the focal elements of different classifiers under an identification frame through a mapping relation established between the entropy value and the uncertain focal elements;
the reliability normalization subunit is used for performing normalization processing on the reliability of the focus elements under different classifiers, so that the basic conditions of the D-S evidence theory are met;
and the D-S evidence theory fusion judgment subunit is used for eliminating conflicts among different classifiers according to the evidence combination rule and acquiring a final fusion judgment result.
8. A confusion behavior conflict management method based on multiple classifiers is characterized by comprising the following steps: the method comprises the following steps:
collecting physiological information and motion information of a measured object;
transmitting and storing physiological information and motion information;
preprocessing physiological information and motion information to obtain input data of a plurality of classifiers, and inputting the preprocessed data into the plurality of classifiers for behavior recognition;
obtaining classification results of a plurality of classifiers, determining whether confusion behaviors exist among the classification results and the types of the confusion behaviors, if so, triggering a conflict detection and elimination task, wherein the conflict detection task carries out marking conversion by obtaining the probability of real labels and behavior information to determine whether conflicts exist, and the elimination task eliminates the conflicts among different classifiers;
the recognition accuracy of the confusion behavior after the conflict is eliminated is calculated.
9. The method according to claim 8, wherein the method for managing confusion behavior conflict based on multiple classifiers comprises: the specific process for determining whether confusion behaviors exist among the classification results and the type of the confusion behaviors comprises the following steps:
and analyzing the confusion matrix obtained by different classifiers, judging whether confusion behaviors exist or not according to a preset threshold value and outputting the type of the confusion behaviors, and if the confusion behaviors exceed the preset threshold value, judging that the confusion behaviors exist.
10. The method according to claim 8, wherein the method for managing confusion behavior conflict based on multiple classifiers comprises: the conflict detection and elimination task process comprises the following steps:
acquiring real labels of behaviors to be identified, determining column indexes of the confused behaviors according to the real labels, and acquiring probability vectors of various behaviors output by different classifiers according to the indexes;
calculating a column index corresponding to the maximum probability value according to the probability vectors of different types of behaviors in different classifiers, and marking correspondingly;
determining whether conflict detection exists between different classifiers according to the marker sequences of the different classifiers, and if not, selecting the marker sequence of any classifier as an output result;
and if the result exists, acquiring entropy values of the uncertain focal elements by different classifiers, selecting a proper mapping relation according to the acquired entropy values to obtain initial reliability distribution of the uncertain focal elements, carrying out normalization processing on the initial reliability of each focal element, carrying out fusion judgment on the normalized result and outputting a final judgment result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011589114.6A CN114758408B (en) | 2020-12-28 | 2020-12-28 | Confusion behavior conflict management system and method based on multiple classifiers |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011589114.6A CN114758408B (en) | 2020-12-28 | 2020-12-28 | Confusion behavior conflict management system and method based on multiple classifiers |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114758408A true CN114758408A (en) | 2022-07-15 |
| CN114758408B CN114758408B (en) | 2024-05-07 |
Family
ID=82324547
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011589114.6A Active CN114758408B (en) | 2020-12-28 | 2020-12-28 | Confusion behavior conflict management system and method based on multiple classifiers |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114758408B (en) |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040236611A1 (en) * | 2003-04-30 | 2004-11-25 | Ge Financial Assurance Holdings, Inc. | System and process for a neural network classification for insurance underwriting suitable for use by an automated system |
| US20050286772A1 (en) * | 2004-06-24 | 2005-12-29 | Lockheed Martin Corporation | Multiple classifier system with voting arbitration |
| CN101639864A (en) * | 2009-08-18 | 2010-02-03 | 东南大学 | Multi-level hierarchical DSmT rapid approximate reasoning fusion method |
| US8032467B1 (en) * | 2008-05-31 | 2011-10-04 | Hrl Laboratories, Llc | Valuation-based learning system |
| CN103093238A (en) * | 2013-01-15 | 2013-05-08 | 江苏大学 | Visual dictionary construction method based on Dempster-Shafer (D-S) evidence theory |
| CN106056163A (en) * | 2016-06-08 | 2016-10-26 | 重庆邮电大学 | Multi-sensor information fusion object identification method |
| CN107368854A (en) * | 2017-07-20 | 2017-11-21 | 华北电力大学（保定） | A kind of circuit breaker failure diagnostic method based on improvement evidence theory |
| CN109540520A (en) * | 2018-11-29 | 2019-03-29 | 中国船舶重工集团海装风电股份有限公司 | A kind of rolling bearing fault fusion diagnosis method based on improvement D-S evidence theory |
| CN110441060A (en) * | 2019-08-07 | 2019-11-12 | 贵州电网有限责任公司 | One kind being based on mechanical failure diagnostic method combined of multi-sensor information |
| CN110826626A (en) * | 2019-11-06 | 2020-02-21 | 西安交通大学 | Confusion matrix-based multi-classifier difference measurement method and system |
| CN110852382A (en) * | 2019-11-12 | 2020-02-28 | 山东大学 | Behavior recognition system based on space-time multi-feature extraction and working method thereof |
| CN111027487A (en) * | 2019-12-11 | 2020-04-17 | 山东大学 | Behavior recognition system, method, medium, and apparatus based on multi-convolution kernel residual network |
| CN111709487A (en) * | 2020-06-22 | 2020-09-25 | 中国科学院空天信息创新研究院 | Underwater multi-source acoustic image substrate classification method and system based on decision-level fusion |
-
2020
- 2020-12-28 CN CN202011589114.6A patent/CN114758408B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040236611A1 (en) * | 2003-04-30 | 2004-11-25 | Ge Financial Assurance Holdings, Inc. | System and process for a neural network classification for insurance underwriting suitable for use by an automated system |
| US20050286772A1 (en) * | 2004-06-24 | 2005-12-29 | Lockheed Martin Corporation | Multiple classifier system with voting arbitration |
| US8032467B1 (en) * | 2008-05-31 | 2011-10-04 | Hrl Laboratories, Llc | Valuation-based learning system |
| CN101639864A (en) * | 2009-08-18 | 2010-02-03 | 东南大学 | Multi-level hierarchical DSmT rapid approximate reasoning fusion method |
| CN103093238A (en) * | 2013-01-15 | 2013-05-08 | 江苏大学 | Visual dictionary construction method based on Dempster-Shafer (D-S) evidence theory |
| CN106056163A (en) * | 2016-06-08 | 2016-10-26 | 重庆邮电大学 | Multi-sensor information fusion object identification method |
| CN107368854A (en) * | 2017-07-20 | 2017-11-21 | 华北电力大学（保定） | A kind of circuit breaker failure diagnostic method based on improvement evidence theory |
| CN109540520A (en) * | 2018-11-29 | 2019-03-29 | 中国船舶重工集团海装风电股份有限公司 | A kind of rolling bearing fault fusion diagnosis method based on improvement D-S evidence theory |
| CN110441060A (en) * | 2019-08-07 | 2019-11-12 | 贵州电网有限责任公司 | One kind being based on mechanical failure diagnostic method combined of multi-sensor information |
| CN110826626A (en) * | 2019-11-06 | 2020-02-21 | 西安交通大学 | Confusion matrix-based multi-classifier difference measurement method and system |
| CN110852382A (en) * | 2019-11-12 | 2020-02-28 | 山东大学 | Behavior recognition system based on space-time multi-feature extraction and working method thereof |
| CN111027487A (en) * | 2019-12-11 | 2020-04-17 | 山东大学 | Behavior recognition system, method, medium, and apparatus based on multi-convolution kernel residual network |
| CN111709487A (en) * | 2020-06-22 | 2020-09-25 | 中国科学院空天信息创新研究院 | Underwater multi-source acoustic image substrate classification method and system based on decision-level fusion |
Non-Patent Citations (8)
| Title |
|---|
| JIE CHEN ET AL.: "Uncertain Information Fusion for Gearbox Fault Diagnosis Based on BP Neural Network and DS Evidence Theory", 2016 12TH WORLD CONGRESS ON INTELLIGENT CONTROL AND AUTOMATION (WCICA), 15 June 2016 (2016-06-15), pages 1372 - 1376, XP032972107, DOI: 10.1109/WCICA.2016.7578248 * |
| RADIM JIROUSEK ET AL.: "A new definition of entropy of belief functions in the Dempster-Shafer theory", INTERNATIONAL JOURNAL OF APPROXIMATE REASONING, 10 October 2017 (2017-10-10), pages 49 - 63 * |
| XIAOZHUAN GAO ET AL.: "Uncertainty measure based on Tsallis entropy in evidence theory", INTERNATIONAL JOURNAL OF INTELLIGENT SYSTEMS, 31 December 2019 (2019-12-31), pages 3105 - 3120 * |
| 康健;李一兵;谢红;林云;: "D-S证据理论与信息熵结合的新算法", 弹箭与制导学报, no. 01, 15 February 2011 (2011-02-15), pages 197 - 200 * |
| 李嘉仪: "基于D-S理论的冲突证据融合算法", 中国优秀硕士学位论文全文数据库 信息科技辑, no. 03, 15 March 2020 (2020-03-15), pages 2 - 5 * |
| 杜保臻: "多源数据融合系统框架与证据理论算法研究及其在智慧医疗中的应用", 中国优秀硕士学位论文全文数据库 信息科技辑, no. 09, 15 September 2019 (2019-09-15) * |
| 贺炎;王科;王忠民;: "用户无关的多分类器融合行为识别模型", 西安邮电大学学报, no. 05, 10 September 2016 (2016-09-10), pages 50 - 54 * |
| 陈超;陈性元;汪永伟;王义功;: "冲突可衡量多源信息融合技术研究", 计算机科学, no. 09, 15 September 2013 (2013-09-15), pages 111 - 115 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114758408B (en) | 2024-05-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111027487B (en) | Behavior recognition system, method, medium and equipment based on multi-convolution kernel residual error network | |
| Musci et al. | Online fall detection using recurrent neural networks on smart wearable devices | |
| JP4590018B1 (en) | Feature value conversion apparatus and feature value conversion method | |
| CN108171278B (en) | Motion pattern recognition method and system based on motion training data | |
| CN106580282A (en) | Human body health monitoring device, system and method | |
| CN114818788A (en) | Tracking target state identification method and device based on millimeter wave perception | |
| KR102280291B1 (en) | Apparatus and method for identify patients with parkinson's disease and patients with podarthritis by performing neural network analysis by various detection information | |
| CN109770913A (en) | A kind of abnormal gait recognition methods based on reverse transmittance nerve network | |
| Kale et al. | Human posture recognition using artificial neural networks | |
| CN110598599A (en) | Method and device for detecting abnormal gait of human body based on Gabor atomic decomposition | |
| CN117690583B (en) | Internet of things-based rehabilitation and nursing interactive management system and method | |
| CN114241270A (en) | Intelligent monitoring method, system and device for home care | |
| Alqarni | Error-less data fusion for posture detection using smart healthcare systems and wearable sensors for patient monitoring | |
| CN114758408B (en) | Confusion behavior conflict management system and method based on multiple classifiers | |
| KR102302719B1 (en) | Apparatus and method for classification of gait type by performing neural network analysis for various detection information | |
| CN116561649A (en) | Diver motion state identification method and system based on multi-source sensor data | |
| CN111403021A (en) | Monitoring method and device | |
| Dong et al. | Real-time physical activity monitoring by data fusion in body sensor networks | |
| Yan et al. | Topological nonlinear analysis of dynamical systems in wearable sensor-based human physical activity inference | |
| CN112784812B (en) | Deep squatting action recognition method | |
| KR20140020440A (en) | Activity diagnosis apparatus and method of the same | |
| Dong et al. | A Hybrid HMM/Kalman filter for tracking hip angle in gait cycle | |
| Khokhlova et al. | Kinematic covariance based abnormal gait detection | |
| İsmail et al. | Human activity recognition based on smartphone sensor data using cnn | |
| Shi et al. | Internet-of-Things Behavior Monitoring System Based on Wearable Inertial Sensors for Classifying Dairy Cattle Health Using Machine Learning |
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 |