CN108426349B - Air conditioner personalized health management method based on complex network and image recognition - Google Patents

Air conditioner personalized health management method based on complex network and image recognition Download PDF

Info

Publication number
CN108426349B
CN108426349B CN201810167495.5A CN201810167495A CN108426349B CN 108426349 B CN108426349 B CN 108426349B CN 201810167495 A CN201810167495 A CN 201810167495A CN 108426349 B CN108426349 B CN 108426349B
Authority
CN
China
Prior art keywords
data
network
user
convolutional neural
deep convolutional
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
Application number
CN201810167495.5A
Other languages
Chinese (zh)
Other versions
CN108426349A (en
Inventor
高忠科
党伟东
侯林华
吕冬梅
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tianjin University
Original Assignee
Tianjin University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tianjin University filed Critical Tianjin University
Priority to CN201810167495.5A priority Critical patent/CN108426349B/en
Publication of CN108426349A publication Critical patent/CN108426349A/en
Application granted granted Critical
Publication of CN108426349B publication Critical patent/CN108426349B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F11/00—Control or safety arrangements
    • F24F11/62—Control or safety arrangements characterised by the type of control or by internal processing, e.g. using fuzzy logic, adaptive control or estimation of values
    • F24F11/63—Electronic processing
    • F24F11/65—Electronic processing for selecting an operating mode
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F11/00—Control or safety arrangements
    • F24F11/50—Control or safety arrangements characterised by user interfaces or communication
    • F24F11/56—Remote control
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F11/00—Control or safety arrangements
    • F24F11/62—Control or safety arrangements characterised by the type of control or by internal processing, e.g. using fuzzy logic, adaptive control or estimation of values
    • F24F11/63—Electronic processing
    • F24F11/64—Electronic processing using pre-stored data
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F2110/00—Control inputs relating to air properties
    • F24F2110/10—Temperature
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F2110/00—Control inputs relating to air properties
    • F24F2110/20—Humidity
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F24—HEATING; RANGES; VENTILATING
    • F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
    • F24F2120/00—Control inputs relating to users or occupants
    • F24F2120/10—Occupancy

Landscapes

  • Engineering & Computer Science (AREA)
  • Signal Processing (AREA)
  • Chemical & Material Sciences (AREA)
  • Combustion & Propulsion (AREA)
  • Mechanical Engineering (AREA)
  • General Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Human Computer Interaction (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)

Abstract

An air conditioner personalized health management method based on complex network and image recognition comprises the following steps: acquiring home environment data, user physiological data and physical examination data; constructing a deep convolutional neural network A to analyze and process the home picture data; at different time intervals, the pulse data X of the user is processed1And user electrocardiographic data X2Carrying out feature extraction to obtain network indexes at different time periods; constructing and training a deep convolutional neural network B, and classifying network indexes at different time intervals by using the deep convolutional neural network B; and adjusting the operation mode of the air conditioner, including automatically adjusting according to the established deep convolutional neural network A and deep convolutional neural network B, the home picture data acquired in real time and the user physiological data, and manually adjusting through a mobile terminal. In the air conditioner regulation and control process, the home environment scene and the physical health condition of the user are combined instead of subjective feeling of the user, so that more accurate temperature and humidity regulation and control can be achieved.

Description

Air conditioner personalized health management method based on complex network and image recognition
Technical Field
The invention relates to an air conditioner management method. In particular to an air conditioner personalized health management method based on complex network and image recognition.
Background
With the development of science and technology, the living standard of people is improved, and the air conditioner becomes an indispensable household appliance in daily life and is more widely applied. The existing control method of the air conditioner is generally remote controller control and key control, the remote controller control cannot play a control effect outside a specific area, and the acceptable range of the air conditioner is limited; the key control needs to be controlled beside the air conditioner, and the key control is difficult to control for wall-mounted air conditioners and other types of air conditioners. And the two methods can not automatically adjust and control, and when indoor activities, personnel and the like change, the control scheme can not be adjusted in time, so that the resources of the machine can not be reasonably distributed, and potential waste can be caused.
Deep learning is used as a new machine learning method, features from samples can be extracted layer by simulating a multilayer structure of human brain processing problems, finally, abstract description of the samples is given, great success is achieved in the aspects of image recognition, natural language processing and the like, and the method is widely applied to various fields. Based on the image recognition of the deep learning, the function improvement can be brought to the air conditioner. Family scenes are captured through the camera, classification is carried out through a deep learning method, a specific control scheme is adopted for each result, and intelligent control is achieved to a certain extent. The complex network provides a new visual angle and a new method for the research of the complex system. Complex networks not only focus on interactions between individuals within the system, but also on the macroscopic characteristics of the system as a whole. The visual graph is used as a new complex network theory, and is widely applied to various fields such as economy, multiphase flow, traffic and the like due to the characteristics of high calculation speed and visual appearance. And the intelligent identification of the home scene is realized by combining deep learning and a complex network theory, and the personalized air conditioner management scheme is customized according to the intelligent identification. Meanwhile, the original control mode can be kept, and the mobile terminal control is supported, so that the use of a user is facilitated.
Disclosure of Invention
The invention aims to solve the technical problem of providing an air conditioner personalized health management method based on complex network and image recognition, which can automatically adjust the operation mode through acquired data such as home environment, user information and the like.
The technical scheme adopted by the invention is as follows: an air conditioner personalized health management method based on complex network and image recognition comprises the following steps:
1) acquiring home environment data, user physiological data and physical examination data;
2) constructing a deep convolutional neural network A to analyze and process the home picture data;
3) combining multiple scales at different time intervalsTheory and visual complex network theory, for user pulse data X1And user electrocardiographic data X2Carrying out feature extraction to obtain network indexes at different time periods;
4) constructing and training a deep convolutional neural network B, and classifying the network indexes of different time periods obtained in the step 3) by using the deep convolutional neural network B;
5) and adjusting the operation mode of the air conditioner, including automatically adjusting according to the established deep convolutional neural network A and deep convolutional neural network B, the home picture data acquired in real time and the user physiological data, and manually adjusting through a mobile terminal.
The step 1) comprises the following steps:
(1) the method comprises the steps that home picture data are obtained in real time through a camera, indoor multipoint temperature data are obtained in real time through a distributed temperature sensor, and indoor multipoint humidity data are obtained in real time through a distributed humidity sensor to form home environment data;
(2) the intelligent bracelet is personalized and customized according to the preference of the user and is used for acquiring the pulse data X of the user in real time1Acquiring user electrocardiogram data X in real time through portable electrocardiogram acquisition equipment2And forming the physiological data of the user.
The construction process of the deep convolutional neural network A in the step 2) comprises the following processes:
(1) acquiring a large amount of home image data, labeling the contents of the home image data, and forming a home image data set by using the number, distribution and home scenes of users in a home environment as tags and the large amount of home image data and corresponding tags;
(2) the method comprises the steps of constructing a deep convolutional neural network A, determining a network structure and parameters to be optimized, taking a formed household picture data set as input, completing feature extraction after multilayer convolution and pooling operation, determining classification errors through a target function, optimizing and updating the parameters to be optimized through back propagation errors, repeatedly training the deep convolutional neural network A until specified times are reached or the classification errors are smaller than a set value, completing classification identification through an output layer, and ensuring accurate identification of the household picture data.
The step 3) comprises the following steps:
(1) for a length L1Of the user's pulse data X1And a length L2User electrocardiogram data X2Respectively carrying out multi-scale transformation, wherein the transformation process is as follows:
the following formulas are adopted to carry out coarse graining on the user pulse data and the user electrocardio data respectively:
Figure BDA0001584878690000021
wherein the content of the first and second substances,
Figure BDA0001584878690000022
is the ith point of the signal obtained after the coarse graining operation,
Figure BDA0001584878690000023
is a signal XaAt the j-th point in (1), XaRepresenting user pulse data X1When a is 2, XaRepresenting user cardiac data X2Corresponding data length LaRespectively take L1、L2β is a scale factor, mu indicates that the data is obtained based on the mean,
Figure BDA0001584878690000024
representing a rounding operation;
(2) respectively calculating coarse-grained variance for the user pulse data and the user electrocardiogram data by adopting the following formula to obtain multi-scale signals:
Figure BDA0001584878690000025
for user pulse data X1And user electrocardiographic data X2Respectively obtaining multi-scale pulse signals
Figure BDA0001584878690000026
And multiscale electrocardiosignals
Figure BDA0001584878690000027
Wherein a is 1
Figure BDA0001584878690000028
Representing a multi-scale pulse signal X1(β)The ith point of (1), when a is 2
Figure BDA0001584878690000029
Representing multiscale cardiac signals X2(β)Point i of (2);
(3) for multi-scale pulse signals X1(β)And multiscale electrocardiosignal X2(β)Respectively analyzing based on a visual complex network theory, and if any two points are in the analysis
Figure BDA00015848786900000210
And
Figure BDA00015848786900000211
at any point in between
Figure BDA00015848786900000212
The conditions are satisfied:
Figure BDA00015848786900000213
then call
Figure BDA00015848786900000214
And
Figure BDA00015848786900000215
is visible to
Figure BDA00015848786900000216
Each point in the network is a network node, and a network connecting edge is determined according to the visibility between the points: if visible, establishing a continuous edge, if invisible, not establishing a continuous edge, and obtaining data XaVisual graph complex network at scale β
Figure BDA0001584878690000031
Scale β is updated to obtain data XaAt a plurality of scalesA plurality of viewable complex networks at a degree;
(4) for each viewable complex network AaExtracting the network average node degree M, the network average node betweenness B, the network average aggregation coefficient C, the network global aggregation coefficient G and the network aggregation coefficient entropy ECAnd network average shortest path SPThe network index of (a);
(5) obtaining a large amount of user pulse data X at different time intervals1And user electrocardiographic data X2And (4) repeating the processes from the step (1) to the step (4) to obtain the network index.
The step 4) comprises the following steps:
(1) the network indexes in different periods are sorted, the health condition grade of the user is determined according to the physical examination data, the health condition grade is used as a label, and the network indexes in different periods and the corresponding label are combined into a network index data set;
(2) determining a network structure and parameters to be optimized of a deep convolutional neural network B, taking a formed network index data set as input, completing input feature extraction after multilayer convolution and pooling operation, determining a classification error through an objective function, optimizing and updating the parameters to be optimized through back propagation errors, repeatedly training the deep convolutional neural network B until specified times are reached or the classification error is less than a set value, completing classification identification through an output layer, ensuring accurate classification of the network index data set, and thus obtaining the grade of the physical health condition of a user.
The automatic adjustment in step 5) comprises:
(1) inputting the home image data acquired in real time into an established deep convolutional neural network A, determining the number and distribution of users in a home environment and a home scene where the users are located, and guiding the adjustment of an air conditioner operation mode by combining temperature data and humidity data;
(2) acquiring network indexes from the user physiological data acquired in real time by adopting the method in the step 3), inputting the network indexes into the established deep convolutional neural network B to obtain the current body health condition grade of the user, and guiding the adjustment of the air conditioner operation mode according to the grade.
The air conditioner personalized health management method based on the complex network and the image recognition has the following beneficial effects:
the deep learning and visual graph complex network theory are combined, the air conditioner is actively regulated and controlled according to a home environment scene and the use history of the user personalized air conditioner, and the influence on the life quality caused by frequent regulation and control of the air conditioner by the user is avoided. In the air conditioner regulation and control process, the home environment scene and the physical health condition of the user are combined instead of subjective feeling of the user, more accurate temperature and humidity regulation and control can be achieved, active care for the user is embodied, and potential energy waste is reduced as much as possible while life experience of the user is improved.
Drawings
FIG. 1 is a flow chart of the air conditioner personalized health management method based on complex network and image recognition according to the present invention;
FIG. 2 is a schematic diagram of the structure of a deep convolutional neural network A (CNN);
fig. 3 is a schematic diagram of a visual networking principle.
Detailed Description
The following describes the air conditioner personalized health management method based on the complex network and the image recognition in detail with reference to the embodiments and the accompanying drawings.
As shown in fig. 1, the air conditioner personalized health management method based on the complex network and the image recognition of the present invention includes the following steps:
1) acquiring home environment data, user physiological data and physical examination data; the method comprises the following steps:
(1) the method comprises the steps that home picture data are obtained in real time through a camera, indoor multipoint temperature data are obtained in real time through a distributed temperature sensor, and indoor multipoint humidity data are obtained in real time through a distributed humidity sensor to form home environment data;
(2) the intelligent bracelet is personalized and customized according to the preference of the user and is used for acquiring the pulse data X of the user in real time1Acquiring user electrocardiogram data X in real time through portable electrocardiogram acquisition equipment2And forming the physiological data of the user.
2) And (3) constructing a deep convolutional neural network A to analyze and process the home picture data, wherein the network structure is shown as figure 2.
The construction process of the deep convolutional neural network A comprises the following processes:
(1) acquiring a large amount of home image data, labeling the contents of the home image data, and forming a home image data set by using the number, distribution and home scenes of users in a home environment as tags and the large amount of home image data and corresponding tags;
(2) the method comprises the steps of constructing a deep convolutional neural network A, determining a network structure and parameters to be optimized, taking a formed household picture data set as input, completing feature extraction after multilayer convolution and pooling operation, determining classification errors through a target function, optimizing and updating the parameters to be optimized through back propagation errors, repeatedly training the deep convolutional neural network A until specified times are reached or the classification errors are smaller than a set value, completing classification identification through an output layer, and ensuring accurate identification of the household picture data. The specific process is as follows:
(2.1) constructing a deep convolutional neural network A according to the following formula: INPUT denotes INPUT, CONV denotes convolutional layer, RELU denotes activation function, POOL denotes pooling layer, question mark? Indicating that the pooling layer is optional, FC indicates a fully-connected layer, where N, M, K all indicate the number of layers and satisfy N ≧ 0 (typically N ≦ 3), M ≧ 0, K ≧ 0 (typically K < 3).
INPUT-＞[[CONV-＞RELU]*N-＞POOL？]*M-＞[FC-＞RELU]*K-＞FC
And (2.2) determining a network structure and parameters to be optimized, wherein the household picture data set is divided into two parts which are respectively used as a training sample set and a test sample set, the deep convolutional neural network A is trained through samples in the training sample set, and the classification error of the deep convolutional neural network A is determined through the samples in the test sample set.
If the input data size is R, the receptive field size of the neurons in the convolutional layer is T, the step length is S, and the number of zero padding is P, then the data size output by the convolutional layer is
Figure BDA0001584878690000041
A pooling layer is periodically inserted between the continuous convolution layers, so that the size of a characteristic graph generated in the middle process is reduced, the data depth is not changed, the parameter scale is reduced, and overfitting can be effectively controlled. Pooling operates independently on each depth slice of the input data, using MAX operations, taking the maximum value within a local region to represent that region. This example uses a 2 x 2 pooling kernel, with 75% of the activation information being discarded. The household picture data is transmitted to the output layer from the input layer through gradual change, and the formula executed by the network is as follows:
Op＝Fn(...(F2(F1(IpW(1))W(2))...)W(n))
wherein, IpRepresenting input data, W(n)Weight lattice representing the nth layer, FnRepresenting the activation function of the n-th layer, OpRepresenting the actual output value.
And (2.3) determining a classification error through an objective function, optimizing and updating the parameter to be optimized through back propagation of the error, and repeatedly training until the specified times are reached or the classification error is small enough, so that a network structure and corresponding parameters which can be used for accurately classifying and identifying the home picture data are determined.
3) Combining a multi-scale theory and a visual complex network theory to perform pulse data X on the user1And user electrocardiographic data X2Carrying out feature extraction to obtain a network index; the method comprises the following steps:
(1) for a length L1Of the user's pulse data X1And a length L2User electrocardiogram data X2Respectively carrying out multi-scale transformation, wherein the transformation process is as follows:
the following formulas are adopted to carry out coarse graining on the user pulse data and the user electrocardio data respectively:
Figure BDA0001584878690000051
wherein the content of the first and second substances,
Figure BDA0001584878690000052
is the ith point of the signal obtained after the coarse graining operation,
Figure BDA0001584878690000053
is a signal XaAt the j-th point in (1), XaRepresenting user pulse data X1When a is 2, XaRepresenting user cardiac data X2Corresponding data length LaRespectively take L1、L2β is a scale factor, mu indicates that the data is obtained based on the mean,
Figure BDA0001584878690000054
representing a rounding operation;
(2) respectively calculating coarse-grained variance for the user pulse data and the user electrocardiogram data by adopting the following formula to obtain multi-scale signals:
Figure BDA0001584878690000055
for user pulse data X1And user electrocardiographic data X2Respectively obtaining multi-scale pulse signals
Figure BDA0001584878690000056
And multiscale electrocardiosignals
Figure BDA0001584878690000057
Wherein a is 1
Figure BDA0001584878690000058
Representing a multi-scale pulse signal X1(β)The ith point of (1), when a is 2
Figure BDA0001584878690000059
Representing multiscale cardiac signals X2(β)Point i of (2);
(3) for multi-scale pulse signals X1(β)And multiscale electrocardiosignal X2(β)Respectively based on visual complex network managementThe analysis is carried out if any two points therein
Figure BDA00015848786900000510
And
Figure BDA00015848786900000511
at any point in between
Figure BDA00015848786900000512
The conditions are satisfied:
Figure BDA00015848786900000513
then call
Figure BDA00015848786900000514
And
Figure BDA00015848786900000515
is visible to
Figure BDA00015848786900000516
Each point in the network is a network node, and a network connecting edge is determined according to the visibility between the points: if visible, establishing a continuous edge, if invisible, not establishing a continuous edge, and obtaining data XaVisual graph complex network at scale β
Figure BDA00015848786900000517
Scale β is updated to obtain data XaA plurality of visual graph complex networks under multiple scales, wherein the visual graph networking principle is shown in FIG. 3;
(4) for each viewable complex network AaExtracting network average node degree M, network average node betweenness B and network average aggregation coefficient
Figure BDA00015848786900000518
Network global aggregation coefficient G and network aggregation coefficient entropy ECAnd network average shortest path SPThe network index of (a); the network index of the visual graph complex network is specifically calculated as follows:
(a) network averaging nodePoint degree M: degree m of any node vvIndicating the number of nodes with edges connecting to the node, then
Figure BDA0001584878690000061
Wherein < > represents the operation of averaging, D represents the total number of network nodes;
(b) network average node betweenness B ═ Bv＞，
Figure BDA0001584878690000062
Represents node betweenness, where σrtNumber of shortest paths, σ, connecting node r and node trt(v) The number of the shortest paths connecting the node r and the node t passing through the node v;
(c) network average aggregation coefficient
Figure BDA0001584878690000063
Representing node aggregation coefficients, wherev,ΔRepresenting the number of closed triangles containing node v in a complex network, τvRepresenting the number of triangles with at least two edges starting from the node v in the complex network;
(d) network global aggregation coefficients
Figure BDA0001584878690000064
(e) Entropy of network aggregation coefficients
Figure BDA0001584878690000065
(f) Network average shortest path
Figure BDA0001584878690000066
Wherein node v and node t are different from each other, and SvtRepresenting the shortest path length between node v and node t.
(5) Obtaining a large amount of user pulse data X at different time intervals1And user electrocardiographic data X2And (4) repeating the processes from the step (1) to the step (4) to obtain the network indexes of different time periods.
4) Constructing and training a deep convolutional neural network B, and analyzing and processing the network indexes obtained in the step 3) by using the deep convolutional neural network B, wherein the method comprises the following steps:
(1) the network indexes in different periods are sorted, the health condition grade of the user is determined according to the physical examination data, the health condition grade is used as a label, and the network indexes in different periods and the corresponding label are combined into a network index data set;
(2) determining a network structure and parameters to be optimized of a deep convolutional neural network B, taking a formed network index data set as input, completing input feature extraction after multilayer convolution and pooling operation, determining a classification error through an objective function, optimizing and updating the parameters to be optimized through back propagation errors, repeatedly training the deep convolutional neural network B until specified times are reached or the classification error is less than a set value, completing classification identification through an output layer, ensuring accurate classification of the network index data set, and thus obtaining the grade of the physical health condition of a user.
5) Adjusting the operation mode of the air conditioner, including automatically adjusting according to the established deep convolutional neural network A, the deep convolutional neural network B, the home picture data acquired in real time and the user physiological data, and manually adjusting through a mobile terminal; for example, when a user wants to adjust the working mode of the air conditioner, the user can directly control the mobile terminal (a mobile phone, a tablet computer and the like) interconnected with the air conditioner, and the adjustment can be quickly and conveniently performed.
The automatic adjustment comprises:
(1) inputting the home image data acquired in real time into an established deep convolutional neural network A, determining the number and distribution of users in a home environment and a home scene where the users are located, and guiding the adjustment of an air conditioner operation mode by combining temperature data and humidity data; the method specifically comprises the following steps:
(1.1) intelligently identifying different users, memorizing the use history and characteristics of the air conditioner operation modes of the different users, and automatically adjusting after confirming the identity of the user;
(1.2) when the number of users changes, adjusting on the premise of ensuring the living experience and energy conservation of each user according to the current environment and the number change; if the number of people increases in winter, the temperature can be properly reduced, and if the number of people decreases, the temperature can be properly increased;
(1.3) when the number of the users is not changed but the number of the users is distributed in a plurality of rooms, adjusting the number of the users in each room according to the current environment on the premise of ensuring the living experience and energy conservation of the users;
(1.4) when the air conditioner starts to work, the system obtains a current scene result through picture data analysis, if a user exercises indoors, the operation mode of the air conditioner can be adjusted, the temperature is enabled to be within the range of 17-22 ℃, the relative humidity is controlled to be about 40%, and the body surface comfort of a human body is kept. The corresponding working mode can be changed when other scenes such as sleeping and working;
and (1.5) after the distribution information of the user is obtained through identification, the air conditioner adjusts the position of the blown air in real time according to the information, and the direct blowing of cold (hot) air to the user is avoided.
(2) Acquiring a network index from the user physiological data acquired in real time by adopting the method in the step 3), inputting the network index into the established deep convolutional neural network B to obtain the current body health condition grade of the user, and guiding the adjustment of an air conditioner operation mode according to the grade; the method specifically comprises the following steps:
(2.1) when the health condition of the user is higher in grade, the air conditioner can properly enlarge the temperature fluctuation range, for example, the temperature fluctuation range can be properly reduced on the basis of the recommended temperature in summer, so that the user has better experience;
and (2.2) when the user health condition is low in grade or large in fluctuation, the air conditioner reduces the temperature fluctuation range and ensures that the user health condition is not deteriorated.
The air conditioner personalized health management method based on the complex network and the image recognition can be integrated into an intelligent home system, is effectively combined with a security system, an intelligent refrigerator system capable of realizing diet recommendation and the like, provides a healthy and comfortable home environment for users, and guarantees the health of the users. If, after the user has taken exercise and rest indoors, the air conditioner can properly reduce the temperature (about 22 ℃) and the humidity (about 35%) aiming at the scene at the moment, the security system improves the corresponding grade, and the refrigerator autonomously searches for corresponding post-exercise recipes to supplement nutrition.
The above description of the present invention and the embodiments is not limited thereto, and the description of the embodiments is only one of the implementation manners of the present invention, and any structure or embodiment similar to the technical solution without inventive design is within the protection scope of the present invention without departing from the inventive spirit of the present invention.

Claims (5)

1. An air conditioner personalized health management method based on a complex network and image recognition is characterized by comprising the following steps:
1) acquiring home environment data, user physiological data and physical examination data;
2) the method comprises the following steps of constructing a deep convolutional neural network A to analyze and process home picture data, wherein the construction process of the deep convolutional neural network A comprises the following processes:
(1) acquiring a large amount of home image data, labeling the contents of the home image data, and forming a home image data set by using the number, distribution and home scenes of users in a home environment as tags and the large amount of home image data and corresponding tags;
(2) constructing a deep convolutional neural network A, determining a network structure and parameters to be optimized, taking a formed household picture data set as input, completing feature extraction after multilayer convolution and pooling operation, determining a classification error through a target function, optimizing and updating the parameters to be optimized through a back propagation error, repeatedly training the deep convolutional neural network A until specified times are reached or the classification error is less than a set value, and completing classification identification through an output layer to ensure accurate identification of the household picture data;
3) combining a multi-scale theory and a visual complex network theory to perform pulse data X on the user at different time intervals1And user electrocardiographic data X2Carrying out feature extraction to obtain network indexes at different time periods;
4) constructing and training a deep convolutional neural network B, and classifying the network indexes of different time periods obtained in the step 3) by using the deep convolutional neural network B;
5) and adjusting the operation mode of the air conditioner, including automatically adjusting according to the established deep convolutional neural network A and deep convolutional neural network B, the home picture data acquired in real time and the user physiological data, and manually adjusting through a mobile terminal.
2. The air conditioner personalized health management method based on the complex network and the image recognition as claimed in claim 1, wherein the step 1) comprises:
(1) the method comprises the steps that home picture data are obtained in real time through a camera, indoor multipoint temperature data are obtained in real time through a distributed temperature sensor, and indoor multipoint humidity data are obtained in real time through a distributed humidity sensor to form home environment data;
(2) the intelligent bracelet is personalized and customized according to the preference of the user and is used for acquiring the pulse data X of the user in real time1Acquiring user electrocardiogram data X in real time through portable electrocardiogram acquisition equipment2And forming the physiological data of the user.
3. The air conditioner personalized health management method based on complex network and image recognition according to claim 1, wherein the step 3) comprises:
(1) for a length L1Of the user's pulse data X1And a length L2User electrocardiogram data X2Respectively carrying out multi-scale transformation, wherein the transformation process is as follows:
the following formulas are adopted to carry out coarse graining on the user pulse data and the user electrocardio data respectively:
Figure FDA0002315858620000011
wherein the content of the first and second substances,
Figure FDA0002315858620000012
is the ith point of the signal obtained after the coarse graining operation,
Figure FDA0002315858620000013
is a signal XaAt the j-th point in (1), XaRepresenting user pulse data X1When a is 2, XaRepresenting user cardiac data X2Corresponding data length LaRespectively take L1、L2β is a scale factor, mu indicates that the data is obtained based on the mean,
Figure FDA0002315858620000021
representing a rounding operation;
(2) respectively calculating coarse-grained variance for the user pulse data and the user electrocardiogram data by adopting the following formula to obtain multi-scale signals:
Figure FDA0002315858620000022
for user pulse data X1And user electrocardiographic data X2Respectively obtaining multi-scale pulse signals
Figure FDA0002315858620000023
And multiscale electrocardiosignals
Figure FDA0002315858620000024
Wherein a is 1
Figure FDA0002315858620000025
Representing a multi-scale pulse signal X1(β)The ith point of (1), when a is 2
Figure FDA0002315858620000026
Representing multiscale cardiac signals X2(β)Point i of (2);
(3) for multi-scale pulse signals X1(β)And multiscale electrocardiosignal X2(β)Respectively analyzing based on a visual complex network theory, and if any two points are in the analysis
Figure FDA0002315858620000027
And
Figure FDA0002315858620000028
at any point in between
Figure FDA0002315858620000029
The conditions are satisfied:
Figure FDA00023158586200000210
then call
Figure FDA00023158586200000211
And
Figure FDA00023158586200000212
is visible to
Figure FDA00023158586200000213
Each point in the network is a network node, and a network connecting edge is determined according to the visibility between the points: if visible, establishing a continuous edge, if invisible, not establishing a continuous edge, and obtaining data XaVisual graph complex network at scale β
Figure FDA00023158586200000214
Scale β is updated to obtain data XaA plurality of visual complex networks at multiple scales;
(4) for each viewable complex network AaExtracting network average node degree M, network average node betweenness B and network average aggregation coefficient
Figure FDA00023158586200000215
Network global aggregation coefficient G and network aggregation coefficient entropy ECAnd network average shortest path SPThe network index of (a);
(5) obtaining a large amount of user pulse data X at different time intervals1And user electrocardiographic data X2And (4) repeating the processes from the step (1) to the step (4) to obtain the network index.
4. The air conditioner personalized health management method based on complex network and image recognition according to claim 1, wherein the step 4) comprises:
(1) the network indexes in different periods are sorted, the health condition grade of the user is determined according to the physical examination data, the health condition grade is used as a label, and the network indexes in different periods and the corresponding label are combined into a network index data set;
(2) determining a network structure and parameters to be optimized of a deep convolutional neural network B, taking a formed network index data set as input, completing input feature extraction after multilayer convolution and pooling operation, determining a classification error through an objective function, optimizing and updating the parameters to be optimized through back propagation errors, repeatedly training the deep convolutional neural network B until specified times are reached or the classification error is less than a set value, completing classification identification through an output layer, ensuring accurate classification of the network index data set, and thus obtaining the grade of the physical health condition of a user.
5. The personalized health management method for air conditioners based on complex network and image recognition as claimed in claim 1, wherein the automatic adjustment of step 5) comprises:
(1) inputting the home image data acquired in real time into an established deep convolutional neural network A, determining the number and distribution of users in a home environment and a home scene where the users are located, and guiding the adjustment of an air conditioner operation mode by combining temperature data and humidity data;
(2) acquiring network indexes from the user physiological data acquired in real time by adopting the method in the step 3), inputting the network indexes into the established deep convolutional neural network B to obtain the current body health condition grade of the user, and guiding the adjustment of the air conditioner operation mode according to the grade.
CN201810167495.5A 2018-02-28 2018-02-28 Air conditioner personalized health management method based on complex network and image recognition Active CN108426349B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810167495.5A CN108426349B (en) 2018-02-28 2018-02-28 Air conditioner personalized health management method based on complex network and image recognition

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810167495.5A CN108426349B (en) 2018-02-28 2018-02-28 Air conditioner personalized health management method based on complex network and image recognition

Publications (2)

Publication Number Publication Date
CN108426349A CN108426349A (en) 2018-08-21
CN108426349B true CN108426349B (en) 2020-04-17

Family

ID=63157255

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810167495.5A Active CN108426349B (en) 2018-02-28 2018-02-28 Air conditioner personalized health management method based on complex network and image recognition

Country Status (1)

Country Link
CN (1) CN108426349B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109140665A (en) * 2018-08-31 2019-01-04 珠海格力电器股份有限公司 Control the method, apparatus and air-conditioning device of air-conditioning
CN109358678B (en) * 2018-11-27 2021-04-13 Oppo广东移动通信有限公司 Electronic device, indoor environment control method and related product
CN112146250B (en) * 2020-09-16 2021-10-15 珠海格力电器股份有限公司 Temperature adjusting method and device, electronic equipment and storage medium

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH038003A (en) * 1989-06-06 1991-01-16 Nissan Motor Co Ltd Automobile air conditioner
CN102563808A (en) * 2012-01-11 2012-07-11 华南理工大学 Automatic control method of indoor environment comfort level
CN104490371A (en) * 2014-12-30 2015-04-08 天津大学 Heat comfort detection method based on physiological parameters of human body
CN104833063A (en) * 2015-06-04 2015-08-12 安徽建筑大学 Air conditioner control method and system
CN105222274A (en) * 2015-09-22 2016-01-06 天津大学 According to the method for operation of air conditioner in human physiological reaction control room
CN106322657A (en) * 2016-08-23 2017-01-11 海信（山东）空调有限公司 Air conditioner control method, air conditioner controller and air conditioner system
CN106503799A (en) * 2016-10-11 2017-03-15 天津大学 Deep learning model and the application in brain status monitoring based on multiple dimensioned network
CN106510619A (en) * 2016-10-11 2017-03-22 天津大学 Electrocardiosignal analysis method based on complicated network and application thereof to intelligent wearing

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH038003A (en) * 1989-06-06 1991-01-16 Nissan Motor Co Ltd Automobile air conditioner
CN102563808A (en) * 2012-01-11 2012-07-11 华南理工大学 Automatic control method of indoor environment comfort level
CN104490371A (en) * 2014-12-30 2015-04-08 天津大学 Heat comfort detection method based on physiological parameters of human body
CN104833063A (en) * 2015-06-04 2015-08-12 安徽建筑大学 Air conditioner control method and system
CN105222274A (en) * 2015-09-22 2016-01-06 天津大学 According to the method for operation of air conditioner in human physiological reaction control room
CN106322657A (en) * 2016-08-23 2017-01-11 海信（山东）空调有限公司 Air conditioner control method, air conditioner controller and air conditioner system
CN106503799A (en) * 2016-10-11 2017-03-15 天津大学 Deep learning model and the application in brain status monitoring based on multiple dimensioned network
CN106510619A (en) * 2016-10-11 2017-03-22 天津大学 Electrocardiosignal analysis method based on complicated network and application thereof to intelligent wearing

Also Published As

Publication number Publication date
CN108426349A (en) 2018-08-21

Similar Documents

Publication Publication Date Title
CN108488987B (en) Control method of air conditioning apparatus, storage medium, and apparatus
JP6925536B2 (en) Systems and methods to control operation
WO2017008321A1 (en) Smart home energy management method based on smart wearable device behavior detection
CN107797459B (en) Control method and device of terminal equipment, storage medium and processor
Ghahramani et al. An online learning approach for quantifying personalized thermal comfort via adaptive stochastic modeling
CN106503623B (en) Facial image age estimation method based on convolutional neural networks
CN108426349B (en) Air conditioner personalized health management method based on complex network and image recognition
Jung et al. Occupant-centered real-time control of indoor temperature using deep learning algorithms
CN112861624A (en) Human body posture detection method, system, storage medium, equipment and terminal
CN108413588B (en) Personalized air conditioner control system and method based on thermal imaging and BP neural network
CN108766528B (en) Diet management system, construction method thereof and food material management method
DE102020213128A1 (en) INFERENCE OF HEAT COMFORT OF A USER USING BODY SHAPE INFORMATION
CN113936335B (en) Intelligent sitting posture reminding method and device
CN107018330A (en) A kind of guidance method and device of taking pictures in real time
CN108446021A (en) Application process of the P300 brain-computer interfaces in smart home based on compressed sensing
CN111465949A (en) Information processing apparatus, information processing method, and program
CN109373499B (en) Air conditioning control method and device
CN111256315A (en) Neural network-based air conditioner control model creation method, control method and air conditioner
WO2022247411A1 (en) Air volume control method and apparatus for working machinery cockpit
CN112364570A (en) Shower comfort optimization method, system and server based on deep residual error network
CN110991478A (en) Method for establishing thermal comfort model and method and system for setting user preference temperature
CN110119723A (en) One kind carries out blocking facial expression recognizing method based on ACNN
CN112734003A (en) Human skin temperature detection method based on deep convolutional network
CN110659594B (en) Thermal comfort attitude estimation method based on AlphaPose
CN115905688B (en) Reference information generation method based on artificial intelligence and intelligent home

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