CN106446876A - Sensing behavior recognition method and device - Google Patents
Sensing behavior recognition method and device Download PDFInfo
- Publication number
- CN106446876A CN106446876A CN201611024234.5A CN201611024234A CN106446876A CN 106446876 A CN106446876 A CN 106446876A CN 201611024234 A CN201611024234 A CN 201611024234A CN 106446876 A CN106446876 A CN 106446876A
- Authority
- CN
- China
- Prior art keywords
- sensing
- matrix
- target information
- data
- information matrix
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
- G06V40/23—Recognition of whole body movements, e.g. for sport training
-
- 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
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention discloses a sensing behavior recognition method and a sensing behavior recognition device. The sensing behavior recognition method comprises the following steps: acquiring sensing information acquired by an infrared sensing device, wherein the infrared sensing device consists of sensor arrays arranged at different acquisition points; according to the sensing information, establishing a target information matrix, wherein the target information matrix is used for showing sensing data determined by the sensor arrays at different time points; according to a deep convolution neural network algorithm, analyzing the target information matrix to determine a sensing behavior. Through the sensing behavior recognition method and the sensing behavior recognition device, small-calculation data amount, non-invasive, low-cost and small-volume body behavior recognition is achieved and the recognition accuracy rate is improved.
Description
Technical field
The present embodiments relate to Activity recognition technology, more particularly, to a kind of sensing Activity recognition method and apparatus.
Background technology
Human bodys' response technology mainly gathers physical activity in certain period of time by the equipment such as camera, sensor
Data message, and the Intelligent Recognition of human body behavior is realized by algorithm.At present, Human bodys' response technology and the intelligence such as smart home
Industry relations can be changed close.The acquisition scheme of human body behavioural information is varied, can be divided into the IMAQ side by camera
Case, wearable sensors acquisition scheme, passive sensor acquisition scheme.
In prior art, the Human bodys' response based on camera information collection, it is desirable to multigroup camera continuous acquisition, counts
Larger according to measuring, relatively costly, and the factor in view of privacy, life staying idle at home invades sense stronger.Wearable sensors are not subject to
The impact of illumination variation, but wearing is inconvenient, and Consumer's Experience is poor.Passive infrared sensor is relative to camera to illumination variation
More robustness, but underaction, poor to more complicated Activity recognition effect.Meanwhile, with geometry logic identification human body row
For method because it is simple, efficiently, adaptability apply more the advantages of good in product market, but be difficult to meet relative complex
Activity recognition demand.Bayes's human body recognition method need data volume little, theoretical improve it is adaptable to sensor human bioequivalence, but
It is poor for complex behavior in continuous time and inenarrable similar Activity recognition effect.The method of convolutional neural networks is known
Other effect preferably, but depends on the data of collection, is commonly used to the computer vision Activity recognition based on camera.
Content of the invention
The present invention provides a kind of sensing Activity recognition method and apparatus it is achieved that calculating that data volume is few, non-intrusion type, cost
Low, small volume Human bodys' response, improves recognition accuracy.
In a first aspect, embodiments providing a kind of sensing Activity recognition method it is characterised in that including:
Obtain the heat transfer agent collecting by infrared sensing device, described infrared sensing device is set by different acquisition point
The sensor array composition put;
Build target information matrix according to described heat transfer agent, described target information matrix is used for representing described sensor array
It is listed in the sensing data of different time points determination;
According to deep layer convolutional neural networks algorithm, described target information matrix is analyzed, determines sensing behavior.
Preferably, described collection point is evenly distributed in physical space, and described sensor array is listed in each collection point
According to predetermined interval setting, described sensor array is made up of six binary system infrared sensors.
It is preferred that described include according to described heat transfer agent structure target information matrix in any of the above-described scheme:
Obtain the correction data of heat transfer agent by difference arithmetic, build target information matrix according to described correction data;
Described target information matrix is three-dimensional matrice, and the row (the 1st dimension) of described three-dimensional matrice is the corresponding sensing data of different time points,
The corresponding sensing data in collection point that row (the 2nd dimension) are different, the number of plies (3-dimensional) is that the sensor array of differing heights collects
Sensing data.
It is preferred that described foundation deep layer convolutional neural networks algorithm is to described target information in any of the above-described scheme
Matrix be analyzed including:
According to formula xj max=max (xji), i=1,2 ..., r × c and xj min=min (xji), i=1,2 ..., r × c
Determine the maxima and minima of data in described objective matrix respectively, wherein j is the number of plies of described objective matrix, r is described
The line number of objective matrix, c is the columns of described objective matrix;
According to formulaDescribed objective matrix is carried out matrixing, is transformed into [- 1,1] empty
Between;
Using the matrix after conversion as convolutional neural networks input data, according to formula
Determine the numerical value of every layer of neutral net, wherein l represents the number of plies of convolutional neural networks, j represents this layer network convolution
The number of plies afterwards, f represents the number of plies before this layer network convolution,For each layer input data, ground floor is the target letter after conversion
Breath matrix, other layers are the output data of last layer,For multiple dimensioned convolution kernel, blFor biasing, random initializtion, σ is activation
Function, T represents Chi Hua.
It is preferred that foundation deep layer convolutional neural networks algorithm is to described target information matrix in any of the above-described scheme
It is analyzed, determine that sensing behavior includes:
Described target information matrix is analyzed according to deep layer convolutional neural networks algorithm, according to analysis result and default
Sensing action determines sensing behavior.
Second aspect, the embodiment of the present invention additionally provides a kind of sensing Activity recognition device, including:
Heat transfer agent acquisition module, for obtaining the heat transfer agent collecting by infrared sensing device, described infrared biography
Induction device is made up of the sensor array arranging in different acquisition point;
Target information matrix deciding module, for building target information matrix, described target letter according to described heat transfer agent
Breath matrix is used for representing the sensing data that described sensor array is listed in different time points determination;
Sensing behavioural analysis determining module, for entering to described target information matrix according to deep layer convolutional neural networks algorithm
Row analysis, determines sensing behavior.
Preferably, described collection point is evenly distributed in physical space, and described sensor array is listed in each collection point
According to predetermined interval setting, described sensor array is made up of six binary system infrared sensors.
In any of the above-described scheme it is preferred that described target information matrix deciding module specifically for:
Obtain the correction data of heat transfer agent by difference arithmetic, build target information matrix according to described correction data;
Described target information matrix is three-dimensional matrice, and the corresponding sensing data of behavior different time points of described three-dimensional matrice is classified as not
With the corresponding sensing data in collection point, the number of plies is the sensing data that collects of sensor array of differing heights.
In any of the above-described scheme it is preferred that described sensing behavioural analysis determining module specifically for:
According to formula xj max=max (xji), i=1,2 ..., r × c and xj min=min (xji), i=1,2 ..., r × c
Determine the maxima and minima of data in described objective matrix respectively, wherein j is the number of plies of described objective matrix, r is described
The line number of objective matrix, c is the columns of described objective matrix;
According to formulaDescribed objective matrix is carried out matrixing, is transformed into [- 1,1] empty
Between;
Using the matrix after conversion as convolutional neural networks input data, according to formula
Determine the numerical value of every layer of neutral net, wherein l represents the number of plies of convolutional neural networks, j represents this layer network convolution
The number of plies afterwards, f represents the number of plies before this layer network convolution,For each layer input data, ground floor is the target letter after conversion
Breath matrix, other layers are the output data of last layer,For multiple dimensioned convolution kernel, blFor biasing, random initializtion, σ is activation
Function, T represents Chi Hua.
In any of the above-described scheme it is preferred that described sensing behavioural analysis determining module specifically for:
Described target information matrix is analyzed according to deep layer convolutional neural networks algorithm, according to analysis result and default
Sensing action determines sensing behavior.
, by obtaining the heat transfer agent collecting by infrared sensing device, described infrared sensing device is not by for the present invention
Sensor array composition with collection point setting;Build target information matrix, described target information square according to described heat transfer agent
Battle array is listed in the sensing data of different time points determination for representing described sensor array;According to deep layer convolutional neural networks algorithm pair
Described target information matrix is analyzed, and determines sensing behavior, solve in prior art identification sensing behavior be relatively costly,
The poor problem of effect, it is achieved that calculating that data volume is few, the Human bodys' response of non-intrusion type, low cost, small volume, improves
Recognition accuracy.
Brief description
The flow chart of the sensing Activity recognition method that Fig. 1 provides for the embodiment of the present invention one;
The position view of the collection point that Fig. 2 provides for the embodiment of the present invention one;
Fig. 3 is the heat transfer agent that one layer of sensor array in infrared sensing device provided in an embodiment of the present invention collects
Schematic diagram;
The convolutional neural networks schematic diagram that Fig. 4 provides for the embodiment of the present invention one;
Fig. 5 extends schematic diagram for the convolution algorithm boundary cycle that the embodiment of the present invention one provides;
Fig. 6 trains schematic diagram for the convolutional neural networks that the embodiment of the present invention one provides;
Fig. 7 trains single output result schematic diagram for the convolutional neural networks that the embodiment of the present invention one provides;
The structure chart of the sensing Activity recognition device that Fig. 8 provides for the embodiment of the present invention two.
Specific embodiment
The present invention is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining the present invention, rather than limitation of the invention.It also should be noted that, in order to just
Part related to the present invention rather than entire infrastructure is illustrate only in description, accompanying drawing.
Embodiment one
The flow chart of the sensing Activity recognition method that Fig. 1 provides for the embodiment of the present invention one, the present embodiment is applicable to right
The situation that human body behavior is identified, the method can be executed by computing device such as computer, specifically includes following steps:
The heat transfer agent that step 101, acquisition are collected by infrared sensing device, described infrared sensing device is by difference
The sensor array composition of collection point setting.
Wherein, infrared sensing device refers to measurement using infrared ray as medium, sensor-based system.In this programme it needs to be determined that
The space of sensing behavior is provided with this infrared sensing device.Exemplary, it is assumed that there are length, width and height are all that 2 meters of physics is empty
Between, now need the behavior that this space is occurred to be identified, can be that the human body behavior to the human body that this space exists is carried out
Identification.Specifically, each edge in this physical space arranges two collection points, the collection that Fig. 2 provides for the embodiment of the present invention one
The position view of point, as shown in Fig. 2 collection point 1 is evenly distributed in physical space to collection point 8.This collection point can basis
The difference of specific physical space and adaptability setting, skilled person will appreciate that, the collection point gathered data obtaining more
More accurate.Wherein, the differing heights at each collection point are provided with sensor array, exemplary, arrange every 25 centimetres
One layer of sensor array, that is, each collection point aforesaid be in differing heights be spaced 25 centimetres be disposed with sensor array.
Each sensor array is made up of binary system infrared sensor, and in this programme, each sensor array can include six two and enter
Sensor processed.Wherein, each sensor array can identify whether detect on 8 directions by this six binary sensors
Sensing data.At each collection point in Fig. 2,8 straight lines of transmitting represent on 8 directions that can detect that.
Specifically, each sensor array passes through whether sensing is detected on 8 directions of six binary sensors identification
Data can take following method:It is assumed that each binary sensor recognizes sensing data, and then set is " 1 ", unidentified to then putting
Position is " 0 ", and because each sensor array comprises six binary sensors, the data that is, each sensor array detects can
It is made up of the binary number of 6, such as " 000111 ", wherein first " 0 " correspond to first binary system in sensor array and passes
The collection result of sensor, second " 0 " correspond to the collection result of second binary sensor in sensor array, class successively
Push away;Before detection, precoding can be carried out for this six bit binary data, different codings correspond to one of 8 directions
Direction, coded system can be encoded using LDPC (Low Density Parity Check Code, low density parity check code),
Exemplary, " 00000 " represents first direction in 8 directions, and " 010010 " represents the second direction in 8 directions,
" 101110 " represent the third direction in 8 directions, the like;When 6 determining in sensor array are binary data
It is determined that this sensor array when consistent with one of corresponding six bit binary data in 8 directions having in precoding
Sensing data is detected in this direction, thus, this infrared sensing device can be adopted to the sensing behavior in physical space
Collection.The heat transfer agent that one layer of sensor array in the infrared sensing device that Fig. 3 provides for the embodiment of the present invention one collects
Schematic diagram, as shown in figure 3, exemplary, the coding of array 2 and array 5 to the direction that should have precoding, then can assert array 2
Detect sensing data with array 5 on default direction.
Step 102, build target information matrix according to described heat transfer agent, described target information matrix is used for representing described
Sensor array is listed in the sensing data of different time points determination.
Wherein, target information matrix is three-dimensional matrice, the behavior different time points corresponding sensing number of described three-dimensional matrice
According to being classified as the corresponding sensing data in different collection points, the number of plies is the sensing data that collects of sensor array of differing heights.
Exemplary, the time interval between different time points can be 0.3s, and that is, target information entry of a matrix element represents a certain moment
The human body target distributed intelligence in certain section of altitude range of a certain node.Optionally, because the restriction of concrete physical space is (as certain section of side
Acquisition node etc. can not be installed in edge), the correction data of heat transfer agent can be obtained by difference arithmetic, according to described correction data structure
Build target information matrix.Specifically, can be that node B is located at a left side of node a it is assumed that node A moves right apart from m to point C
Side, and B is n away from A distance, then can obtain the data of point B (ancestor node before not moving) according to data a, c at point A and C
Optionally, in order to improve the accuracy of infrared sensing device gathered data, each junction sensor array acquisition can be sought
The sum that 6 data arriving change in adjacent time, a sensor array of such as a certain node is listed in the number of quarter in some time collection
According to for " 110010 ", it is assumed that it is " 001011 " that this sensor array is listed in upper time data, then change under current time has 4
Position, counts the data that each node (each node comprises 8 sensor arrays) collects successively and changes in total period
Total bit, and sort successively according to size, sequence number is subtracted one as this node data a high position, the such as node of the example above becomes
The total bit changed comes the 4th, then the data of the sensor array of this moment the example above is changed into " 011 110010 ", due to this
Data embodies the timing of sensor array collection, reduces measure error.
Step 103, foundation deep layer convolutional neural networks algorithm are analyzed to described target information matrix, determine sensing row
For.
Exemplary, using 3 layers of convolutional network, every layer of convolution nuclear structure is as shown in figure 4, Fig. 4 is the embodiment of the present invention one
The convolutional neural networks schematic diagram providing, wherein, maximum convolution kernel size kernelsize=3.Fig. 5 is the embodiment of the present invention one
The convolution algorithm boundary cycle extension schematic diagram providing, is illustrated in figure 5 the signal that convolution algorithm boundary cycle extends pad=1
Result.Optionally, according to formula xj max=max (xji), i=1,2 ..., r × c and xj min=min (xji), i=1,2 ...,
R × c determines the maxima and minima of data in described objective matrix respectively, and wherein j is the number of plies of described objective matrix, and r is
The line number of described objective matrix, c is the columns of described objective matrix;
According to formulaDescribed objective matrix is carried out matrixing, is transformed into [- 1,1] empty
Between;
Using the matrix after conversion as convolutional neural networks input data, according to formula
Determine the numerical value of every layer of neutral net, wherein l represents the number of plies of convolutional neural networks, j represents this layer network convolution
The number of plies afterwards, f represents the number of plies before this layer network convolution,For each layer input data, ground floor is the target letter after conversion
Breath matrix, other layers are the output data of last layer,For multiple dimensioned convolution kernel, blFor biasing, random initializtion, σ is activation
Function, T represents Chi Hua.
By to each convolutional layer result of calculation h in this method(l)Enter row interpolation to realize, in the training of convolutional neural networks
During, training convolutional layer 1 obtains 8*8*32 matrix first, and training pool layer 1 obtains the matrix of 6*6*32, then in same application
State formula training convolutional layer 2 and pond layer 2, obtain the matrix of 4*4*64, then also pass through convolutional layer 3 and pond layer 3 obtains 2*
The matrix of 2*32, then trains full articulamentum, and full articulamentum is similar to convolutional layer, and kernelsize is changed to 1, pad=0, then adopts
Seek training error with softmax-Loss loss function, then error back propagation is updatedAnd bl, iteration training
8000 times,And blConvergence is constant.Fig. 6 trains schematic diagram for the convolutional neural networks that the embodiment of the present invention one provides, as Fig. 6 institute
Show it is assumed that the 8*8 matrix on the left side is divided by often big lattice 2*2, that can be divided into 4*4 big lattice Aij, i represents line number, and j represents row
Number, then shown in figure, according to A11、A12、A21、A22This four big lattice obtain the right upper left 4 little lattice b of matrix11、b12、
b21、b22, the like, according to A12、A13、A22、A23Obtain b13、b14、b23、b24, until according to A33、A34、A43、A44Obtain b55、
b56、b65、b66.
In this step, according to deep layer convolutional neural networks algorithm, described target information matrix is analyzed, determines sensing
Behavior includes:According to deep layer convolutional neural networks algorithm, described target information matrix is analyzed, according to analysis result with advance
If sensing action determines sensing behavior.Exemplary, the setting of default sensing action can be carried out according to the classification of physical activity, move
Work can be tie the shoelace, sitting posture is stretched oneself, jog, polishing-shoes, sitting posture chest expanding exercise, original place jete etc..
Specifically, input data is:According to formula8*8*8 target information square after conversion
Battle array I.
1) convolutional layer 1:32 convolution kernel K are initialized by xavier methodj, using the border extending method shown in Fig. 5, by
Following formula draw 32 corresponding 8*8 characteristic pattern c(1).
Wherein, j is 1,2 ..., 32;F is 1,2 ...., 8.
2) Tanh layer 1:Nonlinear change is carried out according to tanh function.
3) pond layer 1:According to Fig. 6,8*8*32 matrixing is 6*6*32 matrix by method.
4) regularization layer 1:According to formulaPond layer output data is transformed to [- 1,
1], in the range of, obtain matrix n(1), x is the element in matrix.
5) convolutional layer 2:64 convolution kernel K are initialized by xavier methodj, using the border extending method shown in Fig. 5, defeated
Enter the 6*6*32 matrix for regularization layer 1 output, draw 64 corresponding 6*6 characteristic pattern c by following formula(2).
Wherein, j is 1,2 ..., 64;F is 1,2 ...., 32.
6) Tanh layer 2:Nonlinear change is carried out according to tanh function.
7) pond layer 2:According to Fig. 6,6*6*64 matrixing is 4*4*64 matrix by method.
8) regularization layer 2:According to formulaPond layer output data is transformed to [- 1,
1], in the range of, obtain matrix n(2), x is the element in matrix.
9) convolutional layer 3:32 convolution kernel K are initialized by xavier methodj, using the border extending method shown in Fig. 5, defeated
Enter the 4*4*64 matrix for regularization layer 2 output, draw 32 corresponding 4*4 characteristic pattern c by following formula(3).
Wherein, j is 1,2 ..., 32;F is 1,2 ...., 64.
10) Tanh layer 3:Nonlinear change is carried out according to tanh function.
11) pond layer 3:According to Fig. 6,4*4*64 matrixing is 2*2*32 matrix by method.
12) regularization layer 3:According to formulaPond layer output data is transformed to [-
1,1], in the range of, obtain matrix n(3), x is the element in matrix.
13) output layer:Behavior classification O of outputi=Σ wx(3), i is classification number, byAsk
The element that softmax loss function is, x(3)ForElement, w be initialization weight.
14) obtain the changing value of each layer parameter and convolution kernel by way of derivation, update each layer parameter.
15) the new data of input and label, repeat 1) to 14) step, train 8000 times, make parameter tend to constant.
By above-mentioned algorithm, final output result is as shown in fig. 7, the convolutional Neural that provides for the embodiment of the present invention one of Fig. 7
Network training single output result schematic diagram, wherein, abscissa is discernible behavior in default 6, exemplary, 1 expression
Tie the shoelace;2 expressions are stretched oneself;3 expressions are jogged;4 expression polishing-shoes;5 expression chest expanding exercises;6 expressions are caprioled.Accordingly, indulge
Coordinate is the identification probability of corresponding classification, and the classification action similarity due to having is higher, and action may be identified as different
Behavior, now takes probability highest time corresponding classification to be the behavior finally identifying, the probability of the 2nd classification is the most as shown in the figure
Height, then the behavior determined by this algorithm " is stretched oneself " action.
The technical scheme of the present embodiment, by obtaining the heat transfer agent collecting by infrared sensing device, described infrared
Sensing device is made up of the sensor array arranging in different acquisition point;Build target information matrix according to described heat transfer agent,
Described target information matrix is used for representing the sensing data that described sensor array is listed in different time points determination;According to deep layer convolution
Neural network algorithm is analyzed to described target information matrix, determines sensing behavior, solves in prior art in identification sensing
Behavior is problem relatively costly, that effect is poor it is achieved that calculating that data volume is few, the human body of non-intrusion type, low cost, small volume
Activity recognition, improves recognition accuracy.
Embodiment two
The structure chart of the sensing Activity recognition device that Fig. 8 provides for the embodiment of the present invention two, specifically includes as follows:
Heat transfer agent acquisition module 201, for obtaining the heat transfer agent collecting by infrared sensing device, described infrared
Sensing device is made up of the sensor array arranging in different acquisition point;
Target information matrix deciding module 202, for building target information matrix, described target according to described heat transfer agent
Information matrix is used for representing the sensing data that described sensor array is listed in different time points determination;
Sensing behavioural analysis determining module 203, for foundation deep layer convolutional neural networks algorithm to described target information square
Battle array is analyzed, and determines sensing behavior.
The technical scheme of the present embodiment, by obtaining the heat transfer agent collecting by infrared sensing device, described infrared
Sensing device is made up of the sensor array arranging in different acquisition point;Build target information matrix according to described heat transfer agent,
Described target information matrix is used for representing the sensing data that described sensor array is listed in different time points determination;According to deep layer convolution
Neural network algorithm is analyzed to described target information matrix, determines sensing behavior, solves in prior art in identification sensing
Behavior is problem relatively costly, that effect is poor it is achieved that calculating that data volume is few, the human body of non-intrusion type, low cost, small volume
Activity recognition, improves recognition accuracy.
On the basis of technique scheme, described collection point is evenly distributed in physical space, described sensor array
Each collection point is arranged according to predetermined interval, described sensor array is made up of six binary system infrared sensors.
On the basis of technique scheme, described target information matrix deciding module specifically for:
Obtain the correction data of heat transfer agent by difference arithmetic, build target information matrix according to described correction data;
Described target information matrix is three-dimensional matrice, and the corresponding sensing data of behavior different time points of described three-dimensional matrice is classified as not
With the corresponding sensing data in collection point, the number of plies is the sensing data that collects of sensor array of differing heights.
On the basis of technique scheme, described sensing behavioural analysis determining module specifically for:
According to formula xj max=max (xji), i=1,2 ..., r × c and xj min=min (xji), i=1,2 ..., r × c
Determine the maxima and minima of data in described objective matrix respectively, wherein j is the number of plies of described objective matrix, r is described
The line number of objective matrix, c is the columns of described objective matrix;
According to formulaDescribed objective matrix is carried out matrixing, is transformed into [- 1,1] empty
Between;
Using the matrix after conversion as convolutional neural networks input data, according to formula
Determine the numerical value of every layer of neutral net, wherein l represents the number of plies of convolutional neural networks, j represents this layer network convolution
The number of plies afterwards, f represents the number of plies before this layer network convolution,For each layer input data, ground floor is the target letter after conversion
Breath matrix, other layers are the output data of last layer,For multiple dimensioned convolution kernel, blFor biasing, random initializtion, σ is activation
Function, T represents Chi Hua.
On the basis of technique scheme, described sensing behavioural analysis determining module specifically for:
Described target information matrix is analyzed according to deep layer convolutional neural networks algorithm, according to analysis result and default
Sensing action determines sensing behavior.
The said goods can perform the method that provided of the embodiment of the present invention one, possess the corresponding functional module of execution method and
Beneficial effect.
Note, above are only presently preferred embodiments of the present invention and institute's application technology principle.It will be appreciated by those skilled in the art that
The invention is not restricted to specific embodiment described here, can carry out for a person skilled in the art various obvious changes,
Readjust and substitute without departing from protection scope of the present invention.Therefore although being carried out to the present invention by above example
It is described in further detail, but the present invention is not limited only to above example, without departing from the inventive concept, also
Other Equivalent embodiments more can be included, and the scope of the present invention is determined by scope of the appended claims.
Claims (10)
1. a kind of sensing Activity recognition method is it is characterised in that include:
Obtain the heat transfer agent that collects by infrared sensing device, described infrared sensing device is by arranging in different acquisition point
Sensor array forms;
Build target information matrix according to described heat transfer agent, described target information matrix is used for representing that described sensor array is listed in
The sensing data that different time points determine;
According to deep layer convolutional neural networks algorithm, described target information matrix is analyzed, determines sensing behavior.
2. method according to claim 1 is it is characterised in that described collection point is evenly distributed in physical space, described
Sensor array is listed in each collection point and arranges according to predetermined interval, and described sensor array is by six binary system infrared sensors
Composition.
3. method according to claim 1 is it is characterised in that described build target information matrix according to described heat transfer agent
Including:
Obtain the correction data of heat transfer agent by difference arithmetic, build target information matrix according to described correction data;Described
Target information matrix is three-dimensional matrice, and the corresponding sensing data of behavior different time points of described three-dimensional matrice is classified as different
The corresponding sensing data in collection point, the number of plies is the sensing data that collects of sensor array of differing heights.
4. the method according to claim 1 or 3 is it is characterised in that described foundation deep layer convolutional neural networks algorithm is to institute
State target information matrix be analyzed including:
According to formula xjmax=max (xji), i=1,2 ..., r × c and xjmin=min (xji), i=1,2 ..., r × c is respectively
Determine the maxima and minima of data in described objective matrix, wherein j is the number of plies of described objective matrix, r is described target
The line number of matrix, c is the columns of described objective matrix;
According to formulaDescribed objective matrix is carried out matrixing, is transformed into [- 1,1] space;
Using the matrix after conversion as convolutional neural networks input data, according to formula
Determine the numerical value of every layer of neutral net, wherein l represents the number of plies of convolutional neural networks, after j represents this layer network convolution
The number of plies, f represents the number of plies before this layer network convolution,For each layer input data, ground floor is the target information square after conversion
Battle array, other layers are the output data of last layer,For multiple dimensioned convolution kernel, blFor biasing, random initializtion, σ is activation primitive,
T represents Chi Hua.
5. method according to claim 4 is it is characterised in that believe to described target according to deep layer convolutional neural networks algorithm
Breath matrix is analyzed, and determines that sensing behavior includes:
According to deep layer convolutional neural networks algorithm, described target information matrix is analyzed, according to analysis result and default sensing
Action determines sensing behavior.
6. a kind of sensing Activity recognition device is it is characterised in that include:
Heat transfer agent acquisition module, for obtaining the heat transfer agent collecting by infrared sensing device, described infrared sensing dress
Put and be made up of the sensor array arranging in different acquisition point;
Target information matrix deciding module, for building target information matrix, described target information square according to described heat transfer agent
Battle array is listed in the sensing data of different time points determination for representing described sensor array;
Sensing behavioural analysis determining module, for carrying out to described target information matrix point according to deep layer convolutional neural networks algorithm
Analysis, determines sensing behavior.
7. device according to claim 6 is it is characterised in that described collection point is evenly distributed in physical space, described
Sensor array is listed in each collection point and arranges according to predetermined interval, and described sensor array is by six binary system infrared sensors
Composition.
8. device according to claim 6 it is characterised in that described target information matrix deciding module specifically for:
Obtain the correction data of heat transfer agent by difference arithmetic, build target information matrix according to described correction data;Described
Target information matrix is three-dimensional matrice, and the corresponding sensing data of behavior different time points of described three-dimensional matrice is classified as different
The corresponding sensing data in collection point, the number of plies is the sensing data that collects of sensor array of differing heights.
9. the device according to claim 6 or 8 it is characterised in that described sensing behavioural analysis determining module specifically for:
According to formula xjmax=max (xji), i=1,2 ..., r × c and xjmin=min (xji), i=1,2 ..., r × c is respectively
Determine the maxima and minima of data in described objective matrix, wherein j is the number of plies of described objective matrix, r is described target
The line number of matrix, c is the columns of described objective matrix;
According to formulaDescribed objective matrix is carried out matrixing, is transformed into [- 1,1] space;
Using the matrix after conversion as convolutional neural networks input data, according to formula
Determine the numerical value of every layer of neutral net, wherein l represents the number of plies of convolutional neural networks, after j represents this layer network convolution
The number of plies, f represents the number of plies before this layer network convolution,For each layer input data, ground floor is the target information square after conversion
Battle array, other layers are the output data of last layer,For multiple dimensioned convolution kernel, blFor biasing, random initializtion, σ is activation primitive,
T represents Chi Hua.
10. device according to claim 9 it is characterised in that described sensing behavioural analysis determining module specifically for:
According to deep layer convolutional neural networks algorithm, described target information matrix is analyzed, according to analysis result and default sensing
Action determines sensing behavior.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611024234.5A CN106446876B (en) | 2016-11-17 | 2016-11-17 | A kind of sensing Activity recognition method and apparatus |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611024234.5A CN106446876B (en) | 2016-11-17 | 2016-11-17 | A kind of sensing Activity recognition method and apparatus |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106446876A true CN106446876A (en) | 2017-02-22 |
CN106446876B CN106446876B (en) | 2019-10-18 |
Family
ID=58220537
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611024234.5A Active CN106446876B (en) | 2016-11-17 | 2016-11-17 | A kind of sensing Activity recognition method and apparatus |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106446876B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107966447A (en) * | 2017-11-14 | 2018-04-27 | 浙江大学 | A kind of Surface Flaw Detection method based on convolutional neural networks |
CN108764059A (en) * | 2018-05-04 | 2018-11-06 | 南京邮电大学 | A kind of Human bodys' response method and system based on neural network |
CN109009013A (en) * | 2018-06-25 | 2018-12-18 | 彭礼烨 | Intelligent bracelet |
CN109768944A (en) * | 2018-12-29 | 2019-05-17 | 苏州联讯仪器有限公司 | A kind of signal modulation identification of code type method based on convolutional neural networks |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105353876A (en) * | 2015-11-09 | 2016-02-24 | 深圳市云腾智慧科技有限公司 | Multi-point light sensation based spatial gesture identification control system and method |
CN105868779A (en) * | 2016-03-28 | 2016-08-17 | 浙江工业大学 | Method for identifying behavior based on feature enhancement and decision fusion |
-
2016
- 2016-11-17 CN CN201611024234.5A patent/CN106446876B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105353876A (en) * | 2015-11-09 | 2016-02-24 | 深圳市云腾智慧科技有限公司 | Multi-point light sensation based spatial gesture identification control system and method |
CN105868779A (en) * | 2016-03-28 | 2016-08-17 | 浙江工业大学 | Method for identifying behavior based on feature enhancement and decision fusion |
Non-Patent Citations (2)
Title |
---|
刘琮,许维胜,***: "时空域深度卷积神经网络及其在行为识别上的应用", 《计算机科学》 * |
徐川龙,顾勤龙,姚明海: "一种基于三维加速度传感器的人体行为识别方法", 《计算机***应用》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107966447A (en) * | 2017-11-14 | 2018-04-27 | 浙江大学 | A kind of Surface Flaw Detection method based on convolutional neural networks |
CN108764059A (en) * | 2018-05-04 | 2018-11-06 | 南京邮电大学 | A kind of Human bodys' response method and system based on neural network |
CN108764059B (en) * | 2018-05-04 | 2021-01-01 | 南京邮电大学 | Human behavior recognition method and system based on neural network |
CN109009013A (en) * | 2018-06-25 | 2018-12-18 | 彭礼烨 | Intelligent bracelet |
CN109768944A (en) * | 2018-12-29 | 2019-05-17 | 苏州联讯仪器有限公司 | A kind of signal modulation identification of code type method based on convolutional neural networks |
Also Published As
Publication number | Publication date |
---|---|
CN106446876B (en) | 2019-10-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Phyo et al. | Deep learning for recognizing human activities using motions of skeletal joints | |
CN106682697A (en) | End-to-end object detection method based on convolutional neural network | |
CN106446876A (en) | Sensing behavior recognition method and device | |
Haque et al. | Towards viewpoint invariant 3d human pose estimation | |
CN107341452A (en) | Human bodys' response method based on quaternary number space-time convolutional neural networks | |
Wang et al. | Road network extraction: A neural-dynamic framework based on deep learning and a finite state machine | |
CN104978580B (en) | A kind of insulator recognition methods for unmanned plane inspection transmission line of electricity | |
CN105447529B (en) | Method and system for detecting clothes and identifying attribute value thereof | |
CN105095833B (en) | For the network establishing method of recognition of face, recognition methods and system | |
CN107657226B (en) | People number estimation method based on deep learning | |
CN109409371A (en) | The system and method for semantic segmentation for image | |
CN107330357A (en) | Vision SLAM closed loop detection methods based on deep neural network | |
CN102360494B (en) | Interactive image segmentation method for multiple foreground targets | |
CN105678284A (en) | Fixed-position human behavior analysis method | |
CN106774314A (en) | A kind of home-services robot paths planning method based on run trace | |
Cai et al. | Fast mining of a network of coevolving time series | |
CN107358293A (en) | A kind of neural network training method and device | |
CN102831427B (en) | Texture feature extraction method fused with visual significance and gray level co-occurrence matrix (GLCM) | |
CN105469041A (en) | Facial point detection system based on multi-task regularization and layer-by-layer supervision neural networ | |
CN110532850B (en) | Fall detection method based on video joint points and hybrid classifier | |
CN109344822B (en) | Scene text detection method based on long-term and short-term memory network | |
CN107423721A (en) | Interactive action detection method, device, storage medium and processor | |
CN105893936A (en) | Behavior identification method base on fusion of HOIRM and local feature | |
CN103034860A (en) | Scale-invariant feature transform (SIFT) based illegal building detection method | |
CN107451565A (en) | A kind of semi-supervised small sample deep learning image model classifying identification method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |