CN108121446B - Exchange method and system - Google Patents
Exchange method and system Download PDFInfo
- Publication number
- CN108121446B CN108121446B CN201711424281.3A CN201711424281A CN108121446B CN 108121446 B CN108121446 B CN 108121446B CN 201711424281 A CN201711424281 A CN 201711424281A CN 108121446 B CN108121446 B CN 108121446B
- Authority
- CN
- China
- Prior art keywords
- target person
- feature
- gait
- predeterminable area
- time
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/017—Gesture based interaction, e.g. based on a set of recognized hand gestures
-
- 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
- G06V40/25—Recognition of walking or running movements, e.g. gait recognition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Multimedia (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention relates to a kind of exchange method and systems.The method includes：The video data for obtaining target person gait and movement in predeterminable area obtains the gait feature of target person in predeterminable area according to the video data of target person gait in the video data；After the synthesis gait feature of gait feature library inquiry to target person, the motion characteristic for entering predeterminable area target person is obtained according to the video data that target person in video data acts, query actions feature database, obtain the corresponding signature of combined action feature, the corresponding relationship of combined action feature and signature is stored in motion characteristic library, machine instruction corresponding with signature is obtained, is interacted according to machine instruction.The gait feature library and motion characteristic library established by deep learning, to improve convenience, safety and the accuracy of present invention interaction.
Description
Technical field
The present invention relates to interaction technique fields, more particularly to a kind of exchange method and a kind of interactive system.
Background technique
With the development of technology, interaction technique is used among every field, and by taking gesture interaction as an example, gesture interaction is
People makes a series of movement by hand, and when allowing machine that can identify the intention of people, however carrying out gesture interaction, interaction personnel must
Palpus close to gestures acquisition device, then makes the gesture of standard, gesture identifying device identifies the gesture that interactive personnel make, holds
The corresponding interaction of row, it is this to interact method using gesture, it is only applicable to personnel and collector is one-to-one is used cooperatively, one
A collector interacts while can not carrying out with more people, and must move to and do hand in the very close prescribed limit of collector
Gesture movement, also results in interactive inconvenience.In addition, interacting using gesture, it is also unable to satisfy wanting for specific occasion safety
It asks.In order to solve interactive safety, occurs the scheme interacted using voice in the prior art, however interactive voice can be by
To the influence of environment, ambient noise is excessive to will affect interactive accuracy.
Summary of the invention
Based on this, it is necessary to which not convenient, the dangerous and inaccurate problem for existing interactive mode provides a kind of friendship
Mutual method and system.
A kind of exchange method, including：
Obtain the video data of target person gait and movement in predeterminable area；
According to the video data of target person gait in the video data, target person is real-time in acquisition predeterminable area
Gait feature；
The synthesis gait feature of the real-time gait feature and target person in gait feature library is compared, comparison passes through it
Afterwards, the video data acted according to target person in the video data obtains and enters the real-time dynamic of predeterminable area target person
Make feature；Wherein, the comprehensive gait feature in the gait feature library is to each time of target person in history video data
Real-time gait feature carries out what deep learning obtained；
By the combined action aspect ratio pair of target person in the real-time action feature and motion characteristic library, comparison passes through it
Afterwards, query actions feature database obtains the corresponding signature of the combined action feature, wherein motion characteristic is stored in library
The corresponding relationship of combined action feature and signature；Combined action is characterized in history video data in the motion characteristic library
Middle each secondary real-time action feature of target person carries out what deep learning obtained；
Machine instruction corresponding with the signature is obtained, is interacted according to the machine instruction.
A kind of interactive system, including：
Video acquisition module, for obtaining the video data of target person gait and movement in predeterminable area；
Gait analysis module obtains preset areas for the video data according to target person gait in the video data
The real-time gait feature of target person in domain；
Motion analysis module, for the synthesis gait of target person in the real-time gait feature and gait feature library is special
Sign compares, and after comparison passes through, according to the video data that target person in the video data acts, obtains and enters predeterminable area
The real-time action feature of target person；Wherein, the comprehensive gait feature in the gait feature library is to history video counts
Carry out what deep learning obtained according to each secondary real-time gait feature of middle target person；
Enquiry module, for by the combined action aspect ratio of target person in the real-time action feature and motion characteristic library
Right, after comparison passes through, query actions feature database obtains the corresponding signature of the combined action feature, wherein movement is special
The corresponding relationship of combined action feature and signature is stored in sign library；Combined action is characterized in pair in the motion characteristic library
Each secondary real-time action feature of target person carries out what deep learning obtained in history video data；
Interactive module is handed over for obtaining machine instruction corresponding with the signature according to the machine instruction
Mutually.
A kind of computer equipment, including memory, processor and be stored on the memory and can be in the processing
The computer program run on device, the processor realize above-mentioned exchange method when executing the computer program.
Modules in above system can only select any group of part of module therein according to concrete application demand
Close application；It can be configured in distributed system locally or remotely and work.
Above-mentioned exchange method and system obtain the mesh according to each secondary real-time gait feature of target person by deep learning
Mark personnel integrate gait feature, establish gait feature library, when carrying out Gait Recognition, more can fast and accurately identify target
Whether the real-time gait feature of personnel is consistent with the synthesis gait feature in gait feature library, equally, for the dynamic of target person
Make the identification of feature, and combined action feature obtained according to each secondary real-time action feature of target person by deep learning,
Motion characteristic library is established, when carrying out the identification of real-time action feature, can also quick and precisely identify the real-time dynamic of target person
Make feature corresponding to the combined action feature in motion characteristic library.In addition, using acquisition in any position people in setting regions
The gait and limbs/body movement video data of member interacts, and personnel are not necessarily to go to the regulation near collector one by one
Required movement is made in range, is kept interaction more convenient, by identifying the gait feature and motion characteristic of target person, was both mentioned
The convenience of height interaction, accuracy, while also improving interactive safety.
Detailed description of the invention
Fig. 1 is the schematic flow chart of exchange method in an embodiment；
Fig. 2 is the schematic diagram of interactive system in an embodiment.
Specific embodiment
It is with reference to the accompanying drawing and preferably real for the effect for further illustrating technological means adopted by the present invention and acquirement
Example is applied, to the technical solution of the embodiment of the present invention, carries out clear and complete description.
Fig. 1 is the schematic flow chart of exchange method in an embodiment, as shown in Figure 1, the step of the method includes：
S101 obtains the video data of target person gait and movement in predeterminable area.
In this step, predeterminable area can be the closed areas such as household rooms, be also possible to the open spaces such as square.Video
Data can be acquired by video acquisition device, specifically, can install a video acquisition device in predeterminable area for adopting
Collect the video data of target person gait and movement.
S102 obtains target person in predeterminable area according to the video data of target person gait in the video data
Real-time gait feature.
In this step, the video data into target person gait each in predeterminable area can be acquired, according to default
Deep learning algorithm the video data of its gait is handled, obtain the real-time gait of each target person in predeterminable area
Feature.
S103 compares the synthesis gait feature of the real-time gait feature and target person in gait feature library, compares
By later, according to the video data that target person in the video data acts, obtaining and entering predeterminable area target person
Real-time action feature；Wherein, the comprehensive gait feature in the gait feature library is to target person in history video data
Each secondary real-time gait feature of member carries out what deep learning obtained.
In this step, there are the synthesis gait feature corresponding to each target person, comprehensive gait in gait feature library
It is characterized in that secondary real-time gait feature each to target person in history video data carries out what deep learning obtained, is interacting
When, it is compared by real-time gait feature and comprehensive gait feature, can quickly identify target person.With using drilling for time
Into, comprehensive gait feature is constantly updated with deep learning, and more comprehensively obtains the personal unique feature of the target person, with
This can preferably identify the identity of target person.
Optionally, it can choose the depth based on CNN (Convolutional neural network, convolutional neural networks)
Learning algorithm is spent, DBNs (Deep neural networks, depth confidence network) is also possible to, it can also be in deep learning
Using capsule network model is based on, with indicating for the detail analysis relationship of preferably simulative neural network external knowledge, in addition, value
Must illustrate, select other models and the deep learning executed or other video data algorithms obtain comprehensive gait feature and
Combined action feature, it is within the scope of the present invention.
For the deep learning of gait feature, when being on the one hand that target person is taken action every time in acquisition video data, row
Walk the side such as the distance between path and collector, angle etc. difference and individual sports speed, acceleration, movement posture
On the other hand the difference in face is the growth for acquiring target person physiological age, physical function variation leads to gait difference, in this step
Suddenly, these differences of deep learning can more comprehensively grasp the target person and walk about the inherent law of behavior, the comprehensive step of continuous updating
State feature so that when all kinds of complicated/empty condition under, can more fast and accurately identify the real-time gait of target person
Feature.
S104, by the combined action aspect ratio pair of target person in real-time action feature and motion characteristic library, comparison passes through
Later, query actions feature database obtains the corresponding signature of the combined action feature, wherein motion characteristic stores in library
There is the corresponding relationship of combined action feature and signature；Combined action is characterized in history video counts in the motion characteristic library
Carry out what deep learning obtained according to each secondary real-time action feature of middle target person.
With the gait feature same principle in S103 step in gait feature library, motion characteristic library stores different target personnel
Combined action feature can fast and accurately identify the real-time action feature of target person when interacting.Movement is special
The identification of sign is also required to be continued the depth for different acquisition angle and acting target person every time after a period of time
It practises, continuous updating combined action feature, with the evolution for using the time, this discrimination can more quickly, accurately, can be more with this
The interaction of good identification target person is intended to.
In addition, in this step, signature is a kind of special label, a signature uniquely corresponds to a machine
Instruction, in motion characteristic library, signature is pre-set, and can be fixed sequence, or meets a set pattern
Rule.
In one embodiment, the step of motion characteristic establishes corresponding relationship with signature can be：In motion characteristic library
Signature arrange in advance, in typing motion characteristic, target person make in a predetermined sequence it is a series of have refer to
The movement for enabling meaning obtains after collecting a series of action video data with instruction meaning of target person through deep learning
To corresponding combined action feature, be successively entered into motion characteristic library, and in order successively it is corresponding with signature foundation close
System, improves the convenience used, what needs to be explained here is that, each target person, can be according to certainly in typing instruction action
Body needs, and makes customized instruction action, without defining a set of unified instruction action.Alternatively, it is also possible to being fixed in advance
The a set of unified instruction action of justice, target person are made before video acquisition device defined according to the instruction action of unified definition
Movement can movement pre-selection by definition and signature then when establishing the corresponding relationship of motion characteristic and signature
Establish corresponding relationship.
S105 obtains machine instruction corresponding with the signature, is interacted according to the machine instruction.
In this step, the relationship of signature and machine instruction can pre-establish, and inquire signature
Afterwards, corresponding machine instruction directly can be sent to corresponding machine.
Optionally, signature can be indicated with binary field, for example, in the server in advance by binary field
0010 is corresponding with air-conditioning instruction is opened, then collecting the corresponding motion characteristic of 0010 field in interaction, just output is opened empty
The instruction of tune.
The technical solution of this implementation can set identity label for target person, pass through depth under the conditions of human intervention
Study, according to target person in history video data each secondary real-time gait feature, obtain the comprehensive gait feature of target person, build
Vertical gait feature library, when carrying out Gait Recognition, can fast and accurately identify target person real-time gait feature whether with
Synthesis gait feature in gait feature library is consistent, and equally, for the identification of the motion characteristic of target person, and is directed to mesh
Each secondary real-time action depths of features study in mark personnel's history video data, obtains its combined action feature, foundation movement is special
Library is levied, when carrying out the identification of real-time action feature, can also quick and precisely identify that the real-time action feature of target person is corresponding
Combined action feature in motion characteristic library.In addition, using acquisition in the setting regions gait of any position personnel and
Limbs/body movement video data interacts, and personnel make finger without being gone in the prescribed limit near away from collector one by one
Fixed movement keeps interaction more convenient, by identifying the gait feature and motion characteristic of target person, had both improved interactive convenience
Property, accuracy, while also improving interactive safety.
In addition, in the present embodiment, in " preparation stage ", video-unit acquires each target person away from collector respectively
The video data of repeated walking process and the target person are away from collector in defined distance, defined route
Limbs/body movement the video data made corresponding to specific instruction is repeated several times in defined distance, defined position, from
It is middle respectively obtain target person synthesis gait feature and combined action feature, and be stored in respectively foundation gait feature library and
Among motion characteristic library；In " application stage ", a video-unit can acquire and determine, what is occurred simultaneously in video data is more
The real-time gait feature and real-time action feature of a target person.Through application after a period of time, target person in database
Comprehensive gait feature and combined action feature experienced after continuous updating, it might even be possible to acquire from a video-unit, tool
In the video data for there are magnanimity personnel, while determining the identity and movement of multiple target persons.
In one embodiment, this video is obtained after the real-time gait feature of target person in acquisition predeterminable area
The different information of the real-time gait feature of target person and comprehensive gait feature in data, according to different information update
Synthesis gait feature in gait feature library；Or, obtaining the real-time action feature and synthesis of target person in this video data
The different information of motion characteristic updates the combined action feature in the motion characteristic library according to the different information.
The technical solution of the present embodiment is divided from the video data of the target person gait and movement that acquire when each interaction
The real-time gait feature and real-time action feature for indescribably taking this interaction of target person, can be as the synthesis in gait feature library
Gait feature carries out the object of combined action depths of features study in deep learning and motion characteristic library, according to this real-time step
The different information of synthesis gait feature in state feature and gait data library updates the synthesis gait feature in gait feature library,
Equally, according to the difference of the combined action feature in this real-time action feature and action database, update action feature database
In combined action feature.The present embodiment has the ability of real-time update comprehensive gait feature and combined action feature, with when
Between evolution, the information content for the relevant action process for obtaining deep learning enriches constantly, and the technical solution of the present embodiment makes
It is more accurate, more quick to obtain subsequent interaction.
In addition, being directed to above-mentioned different information, identification accuracy effect can be updated by different information
The combined action feature in synthesis gait feature and motion characteristic library in gait feature library.The present embodiment is substantially to hand over every time
It mutually can promote the accuracy of comprehensive gait feature and identification combined action feature next time.
In one embodiment, it obtains in predeterminable area after the real-time gait feature of target person, it can also be according in step
Whether with the difference of the real-time gait feature of target person synthesis gait feature within a preset range is inquired in state feature database,
Determine whether the target person has the permission into predeterminable area；If it is not, then exporting warning message.
It, can be by whether there is the real-time step into the target person of predeterminable area in gait feature library in the present embodiment
State feature, if existing standard is that comprehensive gait is special from the real-time gait feature and gait data library extracted in video data
Whether the difference of sign is within the scope of preset, if so, determining that the target person can enter predeterminable area.
Optionally, for predeterminable area, the target person of the comprehensive gait feature of typing not in gait feature library
It is the equal of " stranger ", then exporting warning message when detecting that " stranger " enters the predeterminable area.Specifically, alarm signal
Breath can be a kind of alarm for embodying danger signal, be also possible to a kind of signal etc. of trigger the server sending preset instructions
Deng.
Further, if there is the synthesis gait feature into the target person of predeterminable area in gait feature library, that
Explanation, which can enter predeterminable area, specifically, the permission into predeterminable area further includes：Lack of competence, part
Permission and complete permission, wherein lack of competence indicates that the permission for not having into predeterminable area, part permission indicate preset time period
Has the permission into predeterminable area, complete permission indicates the full-time permission having into predeterminable area；If it is determined that the mesh
Mark personnel have the permission into the predeterminable area, then detect to the permission into predeterminable area of the target person,
If detecting that the target person is that lack of competence target person or part permission personnel are default in the entrance of non-default period
Region then exports warning message, the alarm of the target person of the permission different from not entering into predeterminable area, in the present embodiment
Alarm can be the signal for issuing suggestive sound and/or trigger the server sending preset instructions.
Optionally, the synthesis gait feature that can preset each target person in gait feature library corresponds to its entrance
The permission of predeterminable area can correspond to obtain target person correspondence in the synthesis gait feature for getting certain target person
The permission into predeterminable area.
It optionally, can also include accompanying permission into the permission of predeterminable area, lack of competence target person appears in default
When region, then warning message is exported, current slot has the permission and/or company into predeterminable area if also detecting at this time
The target person of permission then stops exporting warning message.
For example the synthesis gait feature of the members (guest including allowing access into family) of one family is complete
At the operation in typing gait feature library, these target persons are imparted phase in initial acquisition gait feature respectively
The permission answered：Wherein host and hostess all have the complete permission into family's all areas；Nurse have into parlor, dining room,
The complete permission in the regions such as kitchen, my bedroom, child room, has the part permission into master bedroom；Children's lack of competence into
Enter kitchen, Ke Renfang, bathroom, laundry, but have accompanied with the male/female owner or nurse for accompanying permission under can be into
Enter bathroom；Guest has the complete permission for entering the regions such as parlor, dining room, guest bedroom, and lack of competence enters master bedroom, children
The regions such as bedroom, nurse bedroom, laundry；For sharing bathroom, sharing the regions such as toilet, when there is target person to initially enter
Later, other target person lacks of competence enter, etc..
The personnel without the comprehensive gait feature of typing except family's members are exactly for this region of family
Stranger is " invader "；Once it was found that " invader " must send out the alarm of highest level.
Above-described embodiment has further refined the permission that target person enters predeterminable area, to further improve this
The accuracy of technical solution in embodiment, kitchen, laundry, bathroom in some scenarios, such as in family etc. have may
It the space of risk factor can alert (report when the target person (children) for collecting not permission enters kitchen
It is alert), but at this time if there is entering kitchen permission and when target person (the parent or nurse) entrance with permission is accompanied,
Meeting stop alarm, to effectively carry out safety management to target person (children).
In order to further realize interactive safety, in one embodiment, can also the target person enter described in
After predeterminable area, the movement track of target person is also detected, by preset action in the movement track and movement track library
Track comparison；Wherein, the corresponding movement track of target person is pre-set in the movement track library；If the two error is pre-
If outside range, exporting warning message.
In the present embodiment, the movement track of each target person, target person can be pre-established in movement track library
Member enters predeterminable area, and when taking action in predeterminable area, if movement track deviates the action rail stored in movement track library
Mark then exports warning message immediately, to remind target person walking in movement track as defined in it.Furthermore it is possible to set in advance
Certain trajector deviation is set, by preset movement track in the movement track of real time contrast's target person and movement track library,
If the trajector deviation of the two is more than preset trajector deviation, determine the two error outside preset range.
Further, movement track is also possible to the regional scope in predeterminable area, realization principle and movement track power
Limit similar, which is not described herein again.
In another embodiment, residence time of the target person in the predeterminable area is also detected, if when the stop
Between be greater than the preset residence time, then export the warning message.
It in some cases, safely may be and its important, using the interaction schemes of the present embodiment, to pass through setting target
Movement track of the personnel in predeterminable area and the residence time in predeterminable area, the safety in the region greatly ensured.
For example nurse enters master bedroom and does cleaning, can specify that its cleaning sequence to master bedroom：Use rag
Clean path of passing through；With dust catcher/mop cleaning path；And it provides to complete clean overall time (specific implementation：
By the video overall process of video acquisition device acquisition hostess explanation and the lower nurse's corresponding operating of guidance, as setting this permission
Basis；It after doing " putting wide range regulation " to path and time again, is stored in movement track library)；If during practical cleaning,
It was found that nurse is longer than the expected regulation in movement track library the residence time near the desk/cabinet for putting casket, then by servicing
Device issues preset prompt, and the hostess is prompted, checked relevant valuables, checked in time afterwards；If
Discovery is not inconsistent in verification, can have access to nurse's details of associated video during cleaning and be screened.
In addition, the permission of interactive instruction is also and its important, carried out in the permission for entering predeterminable area to target person
After verifying, it is also necessary to verify the permission of the machine instruction of target person.Specifically, inquiring in the motion characteristic library and including
When the combined action feature of target person, the corresponding signature of the combined action feature is obtained.For example, children do not enter into
The right in kitchen, the right being not turned on regulating gas stove can open live signal language when finding that it enters kitchen
Sound, it is desirable that it leaves, but if it find that its not only without departing from, issue the instruction that the gas furnace intensity of a fire increases instead, then can be with
Field alert is issued, fuel gas supply valve is simultaneously closed off；Server can be simultaneously emitted by corresponding predetermined warning.
Optionally, if the permission of the not no corresponding machine instruction of the motion characteristic of target person, can export alarm command,
It is alarmed according to alarm command.For uniform machinery, different people possesses different operating rights, therefore, is corresponding with difference
Machine instruction permission, by each target person being arranged the permission of corresponding machine instruction, to guarantee target person
Its instruction having permission can only be exercised, to improve the safety of the present embodiment exchange method.
In one embodiment, multiple target persons be may be simultaneously present in predeterminable area, if multiple target persons are done simultaneously
The identical movement of instruction meaning out, then system can extract multiple real-time action features, and these real-time action features are equal
When the adjusting of corresponding uniform machinery, machine can only respond one of instruction, in order to solve this problem, specifically, can be with
Each target person is set for the precedence information of the machine instruction of uniform machinery, there are multiple targets in the predeterminable area
When personnel, two of them or the above persons are directed to uniform machinery simultaneously and operate, only in response to the dynamic of highest priority personnel
It instructs.
Multiple target persons are existed simultaneously in the present embodiment, in predeterminable area, and priority is only extracted from video data
Highest corresponds to the real-time action feature of target person, therefore, excellent for other to respond the instruction of highest priority target person
The first lower target person of grade does not extract it and instructs real-time action feature, to reduce the probability of " non-productive work " appearance；Separately
Outside, the limbs/body movement made unconscious for certain target persons for not having permission, machine will not also be known
Not, to avoid meaningless task.
For the deep learning part in the present invention, when specific operation, target person is often in following different shape
State：The conditions such as the distance between motion path and collector, angle and displacement dynamics/speed, athletic posture etc. can
Difference, if, when above-mentioned movement difference occurs in target person, cannot accurately be identified without deep learning,
The present invention uses deep learning, finds the inherent law of target person gait feature difference, is adjusted to comprehensive gait feature,
The accuracy for promoting identification can identify that there are above-mentioned the difference even gaits of other differences, to mention when interacting next time
The accuracy and foresight of height interaction.
Deep learning can also be according to the rule of the movement disparity range of real-time action feature when each interaction, to comprehensive dynamic
It is adjusted as feature, when interacting next time, even if the movement of target person is not standard and consistent, as long as according to movement
The rule of disparity range can identify above-mentioned inconsistent, non-type movement, to interact.
In addition, with the variation at target person age, in fact it could happen that movement difference physiologically is understood by deep learning
The rule of its change of divergence can accurately predict combined action feature and comprehensive gait that target person is likely to occur variation
Feature.
In an optional embodiment, the available target person of deep learning is with the combined action feature after change of age
And the rule of comprehensive gait feature, within certain period after target person is wandered away, this system can pass through deep learning
Inference is inferred to synthesis gait feature of the target person after the period, so that lost target people still can be found
Member.
In another embodiment, several interactive information in the historical data are obtained, wherein every interactive information is extremely
Include less：Comprehensive gait feature, combined action feature, the corresponding machine instruction of combined action feature, spatial information, time letter
The information such as breath, environmental factor information, manipulation result；Big data analysis is carried out for above- mentioned information in history interaction, obtains target
The operation preference information of personnel；By the synthesis gait feature ratio of target person in the real-time gait feature and gait feature library
It is right, after comparison passes through, further include：According to the corresponding machine instruction of operation preference information of the target person, transfer in advance
Machine instruction interacts.
The technical solution of the present embodiment, time, environment occurred by big data analysis each usage record etc. are all kinds of
Relevant information, to obtain the operation preference information of target person, operation preference information is indicated, when target person is in certain environment
When certain time appears in target area, the combined action feature that target person most probable executes is certainly, inclined in order to improve operation
The accuracy of good information prediction, can be with the difference of the real-time action feature of every usage record of Accurate Analysis, to be based on this
A little differences and difference, the accurate movement predicting target person and may executing, advanced optimize exchange method, improve this implementation
The accuracy and foresight of example exchange method.
Hereinafter, further illustrating technical solution of the present invention in conjunction with a specific embodiment.
By taking specific family as an example, kinsfolk includes two parents, a child and a nurse, the electric appliance of the family
Including water heater, television set, refrigerator, computer, crouch indoor lamp etc..There are 3 bedrooms, 1 kitchen, 1 parlor, 2 bathrooms, 2
A toilet is independent predeterminable area respectively.
The video data of two parent's gaits of acquisition, and the video data of acquisition instructions movement in advance first, respectively obtains
The synthesis gait feature and combined action feature of two parents, two parents possess whole interaction permissions, therefore, two parents
The operational order for thering is combined action feature to correspond to all electric appliances.
Followed by child, for child, there may be potential danger for water heater, if opening accidentally,
Unfavorable consequence may be brought, therefore, child is not turned on the permission of water heater.It is long for the child that can operate computer
Time also brings along certain harm using computer, therefore the computation permission of child is only set as the several small of specified time period
When.Tv volume can generate noise effect to other spaces of family, generate damage to the hearing of Field Force, and child is to TV
The adjusting permission of machine volume is provided with the upper limit, and (after reaching upper limit volume, the instruction action that volume is promoted to it is no longer rung
It answers).Corresponding permission can also be arranged to child in the unlatching of other electric appliances or the adjusting of parameter.
For nurse, mainly enter the different rights in 3 bedrooms, nurse has disengaging child room and oneself room
Between permission, but for the room of parent, the permission of corresponding period is set, i.e., can enter householder's room in the corresponding period
Between, in addition, also setting up movement track and residence time of the nurse between householder's room, before the deadline, nurse enters house
Clean-up operation is simultaneously completed according to pre-set movement track in long room, leaves in the defined time, if discovery and regulation are not
When according with plot, then server warning note is carried out in time.
After setting above-mentioned permission, a special scenes is selected to be described as follows：Child enters bathroom, it is intended to open
Water heater at this point, detecting child's not permission, and passes through the identification of child's identity, alarms, one side server mentions
Wake up parent or nurse, and on the other hand warning child in scene leaves, if nurse or parent enter bathroom at this time, stops report at this time
It is alert.Child attempted to open computer in the non-permission time, was child by the verifying to identity, and detect current slot child and do not have
There is out the permission of computer, computer is not responding to the operational order of child.
Application for big data, male owner is after next go back home, and convention can change one's clothes, have a bath, through after a period of time
After instructing light, air conditioner, water heater to adjust the big data analysis of design parameter for it, when again identifying that under male owner
After class's time goes home, the air conditioner of master bedroom is automatically turned on, and temperature is consistent with the usual requirement of male owner；Bedroom and Bathroom
Light automatic opening when male owner enters relevant range, and lamp illuminance and tone comply fully with male owner and usually want
It asks；The water temperature of water heater is consistent with the water temperature that male owner is usually adjusted etc..
Based on thought identical with the exchange method in above-described embodiment, the present invention also provides interactive systems, which can
For executing above-mentioned exchange method.For ease of description, it in the structural schematic diagram of interactive system embodiment, illustrate only and this
It the relevant part of inventive embodiments can be with it will be understood by those skilled in the art that the restriction of schematic structure not structure paired systems
Including perhaps combining certain components or different component layouts than illustrating more or fewer components.
Fig. 2 is the schematic diagram of interactive system in an embodiment, as shown in Fig. 2, the system comprises：
Video acquisition module 201, for obtaining the video data of target person gait and movement in predeterminable area.
Gait analysis module 202 obtains default for the video data according to target person gait in the video data
The real-time gait feature of target person in region.
Motion analysis module 203, for walking the comprehensive of target person in the real-time gait feature and gait feature library
State aspect ratio pair after comparison passes through, according to the video data that target person in the video data acts, is obtained to enter and be preset
The real-time action feature of regional aim personnel；Wherein, the comprehensive gait feature in the gait feature library is regarded to history
Each secondary real-time gait feature of frequency target person in carries out what deep learning obtained.
Enquiry module 204, for the combined action of target person in the real-time action feature and motion characteristic library is special
Sign compares, and after comparison passes through, query actions feature database obtains the corresponding signature of the combined action feature, wherein dynamic
Make the corresponding relationship that combined action feature and signature are stored in feature database；Combined action feature in the motion characteristic library
It is that secondary real-time action feature progress deep learning each to target person in history video data obtains.
Interactive module 205 is carried out for obtaining machine instruction corresponding with the signature according to the machine instruction
Interaction.
The embodiment of above-mentioned interactive system, above-mentioned exchange method and system, by deep learning, according to target person history
Each secondary real-time gait feature in video data obtains the comprehensive gait feature of target person, establishes gait feature library, carrying out gait
When identification, it more can fast and accurately identify whether the real-time gait feature of target person walks with the synthesis in gait feature library
State feature is consistent, and equally, for the identification of the motion characteristic of target person, and by deep learning, is gone through according to target person
Each secondary real-time action feature, obtains combined action feature, establishes motion characteristic library in history video data, special carrying out real-time action
When sign identification, it is dynamic can also quick and precisely to identify that the real-time action feature of target person corresponds to the synthesis in motion characteristic library
Make feature.In addition, using acquisition in the setting regions gait of any position personnel and limbs/body movement video data into
Row interaction, personnel make required movement without being gone in the prescribed limit near away from collector one by one, keep interaction more convenient, lead to
The gait feature and motion characteristic for crossing identification target person, had both improved interactive convenience, accuracy, while also improving friendship
Mutual safety.
In one embodiment, further include update module, obtain the real-time gait feature of target person in this video data
With the different information of comprehensive gait feature, updated in the gait feature library according to the result to the different information deep learning
Synthesis gait feature；It is also used to obtain the real-time action feature and combined action feature of target person in this video data
Different information updates the combined action feature in the motion characteristic library according to the result to the different information deep learning.
It in one embodiment, further include authority management module, for according to whether being inquired in gait feature library and mesh
The synthesis gait feature of the difference of the real-time gait feature of mark personnel within a preset range, determines whether the target person has
Into the permission of predeterminable area；If it is not, then exporting warning message.
Optionally, the permission into predeterminable area further includes：Lack of competence, part permission and complete permission, wherein nothing
Permission indicates that the permission for not having into predeterminable area, part permission indicate that preset time period has the power into predeterminable area
Limit, complete permission indicate the full-time permission having into predeterminable area；Enquiry module 203 is also used to obtain into predeterminable area
The permission into predeterminable area of target person；Detect that the lack of competence target person enters predeterminable area, or detection
To part permission personnel when the non-default period entering predeterminable area, warning message is exported.
Specifically, if enquiry module 203 is also used to also detect that current slot has into the permission of predeterminable area
Target person then stops exporting warning message.
It in one embodiment, further include movement track contrast module, the movement track contrast module is used in the mesh
Mark personnel enter after the predeterminable area, the movement track of target person are also detected, by the movement track and movement track
Preset movement track comparison in library；Wherein, the corresponding movement track of target person is pre-set in the movement track library；
If the two error outside preset range, exports warning message.
It in another embodiment, further include time contrast module, for detecting target person in the predeterminable area
Residence time exports the warning message if the residence time is greater than the preset residence time.
In another optional embodiment, the enquiry module is also used to inquire in the motion characteristic library comprising described
When combined action feature, the corresponding signature of the combined action feature is obtained.
It in one embodiment, further include interaction habits prediction module, the interaction habits prediction module is described for obtaining
Several interactive information in historical data, wherein every interactive information includes at least：Comprehensive gait feature, combined action are special
Sign, the corresponding machine instruction of combined action feature, spatial information, temporal information and environmental factor information；According to every interaction
The spatial information that records in information, temporal information, environmental factor count greatly to this interactive information and history mutual information
According to analysis, the operation preference information of target person is obtained；Interaction habits prediction module is also used to the behaviour according to the target person
Make the corresponding machine instruction of preference information, transfers machine instruction in advance and interact.
It optionally, further include priority block, the priority block is for detecting target person in the predeterminable area
Precedence information, the machine instruction of the high target person of preferential answering priority；Wherein, each target person needle is preset
To the precedence information of the machine instruction of uniform machinery.
Modules in above system can only select any group of part of module therein according to concrete application demand
Close application；It can be configured in distributed system locally or remotely and work.
It will appreciated by the skilled person that realizing all or part of the process in above-described embodiment method, being can
To be completed by the relevant hardware of computer program instructions, each generic module be can be respectively independent or whole
Combination or part are composed, and can only select part of application according to concrete application demand and can in any combination
Hardware；The program can be stored in computer-readable storage medium, sells or uses as independent product.The journey
When being executed, all or part of the steps of the embodiment such as above-mentioned each method can be performed in sequence.Wherein, the storage medium can be
Magnetic disk, CD, read-only memory (Read-Only Memory, ROM) or random access memory (Random Access
Memory, RAM) etc..
Each technical characteristic of embodiment described above can be combined arbitrarily, for simplicity of description, not to above-mentioned reality
It applies all possible combination of each technical characteristic in example and its corresponding application is all described, as long as however, these technologies
Contradiction is not present in the combination of feature, all should be considered as described in this specification.
The embodiments described above only express several embodiments of the present invention, and the description thereof is more specific and detailed, but simultaneously
It cannot therefore be construed as limiting the scope of the patent.It should be pointed out that coming for those of ordinary skill in the art
It says, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to protection of the invention
Range.Therefore, the scope of protection of the patent of the invention shall be subject to the appended claims.
Claims (12)
1. a kind of exchange method, which is characterized in that including：
Obtain the video data of target person gait and movement in predeterminable area；
According to the video data of target person gait in the video data, the real-time gait of target person in predeterminable area is obtained
Feature；
The synthesis gait feature of the real-time gait feature and target person in gait feature library is compared, after comparison passes through,
According to the video data that target person in the video data acts, the real-time action spy for entering predeterminable area target person is obtained
Sign；Wherein, the comprehensive gait feature in the gait feature library is real-time to target person each time in history video data
Gait feature carries out what deep learning obtained；
By the combined action aspect ratio pair of target person in the real-time action feature and motion characteristic library, after comparison passes through,
Query actions feature database obtains the corresponding signature of the combined action feature, wherein motion characteristic is stored with synthesis in library
The corresponding relationship of motion characteristic and signature；Combined action is characterized in mesh in history video data in the motion characteristic library
Each secondary real-time action feature of mark personnel carries out what deep learning obtained；
Machine instruction corresponding with the signature is obtained, is interacted according to the machine instruction.
2. exchange method according to claim 1, which is characterized in that obtain the real-time gait of target person in predeterminable area
The step of after feature, including：
The real-time gait feature of target person in this video data and the different information of comprehensive gait feature are obtained, according to described
Different information updates the synthesis gait feature in the gait feature library；
Or, further including the step of obtaining after the corresponding signature of the combined action feature：
The real-time action feature of target person in this video data and the different information of combined action feature are obtained, according to described
Different information updates the combined action feature in the motion characteristic library.
3. exchange method according to claim 1, which is characterized in that obtain the real-time gait of target person in predeterminable area
The step of after feature further includes：
According to whether being inquired in gait feature library with the difference of the real-time gait feature of target person within a preset range
Comprehensive gait feature, determines whether the target person has the permission into predeterminable area；
If it is not, then exporting warning message.
4. exchange method according to claim 3, which is characterized in that the permission into predeterminable area includes：Have no right
Limit, part permission, and/or complete permission, wherein lack of competence indicates the permission for not having into predeterminable area, part authority list
Show that preset time period has the permission into predeterminable area, complete permission indicates the full-time permission having into predeterminable area；
The step of after determining that the target person has permission into predeterminable area, including：
Obtain the Permission Levels into predeterminable area of target person in predeterminable area；
Detect that the lack of competence target person enters predeterminable area, or detect part permission personnel the non-default period into
When entering predeterminable area, warning message is exported.
5. exchange method according to claim 4, which is characterized in that the permission into predeterminable area further includes：It accompanies
It is described that permission is accompanied to indicate the permission for entering predeterminable area simultaneously with other target persons with permission；
It detects that the lack of competence target person enters predeterminable area described, or detects part permission personnel when non-default
Between section when entering predeterminable area further include the step of generation after alarm command：
If also detecting target person of the current slot with predeterminable area permission is entered and/or with the company permission,
Then stop exporting warning message.
6. exchange method according to claim 1, which is characterized in that further include：
After the target person enters the predeterminable area, the movement track of target person is also detected, by the action rail
Preset movement track compares in mark and movement track library；Wherein, target person is pre-set in the movement track library
Movement track；
If the two error outside preset range, exports warning message.
7. exchange method according to claim 6, which is characterized in that further include：
Residence time of the target person in the predeterminable area is detected, if the residence time is greater than the preset residence time,
Then export warning message.
8. exchange method according to claim 1, which is characterized in that the query actions feature database obtains the synthesis
The step of motion characteristic corresponding signature, including：
When inquiring in the motion characteristic library comprising the combined action feature, the corresponding spy of the combined action feature is obtained
Sign label.
9. exchange method according to claim 1-8, which is characterized in that further include：
The history mutual information in historical data is obtained, wherein the interactive information includes at least：Comprehensive gait feature, synthesis are dynamic
Make feature, the corresponding machine instruction of combined action feature, spatial information, temporal information and environmental factor information；
According to recorded in every interactive information spatial information, temporal information, environmental factor, to this interactive information and history
Interactive information carries out big data analysis, obtains the operation preference information of target person；
It is compared by the synthesis gait feature of the real-time gait feature and target person in gait feature library, comparison passes through it
Afterwards, further include：
According to the corresponding machine instruction of operation preference information of the target person, machine instruction is transferred in advance and is interacted.
10. exchange method according to claim 1-8, which is characterized in that further include：
The precedence information for detecting target person in the predeterminable area, the machine for responding the target person of highest priority refer to
It enables；Wherein, each target person is preset for the precedence information of the machine instruction of uniform machinery.
11. a kind of interactive system, which is characterized in that including：
Video acquisition module, for obtaining the video data of target person gait and movement in predeterminable area；
Gait analysis module obtains in predeterminable area for the video data according to target person gait in the video data
The real-time gait feature of target person；
Motion analysis module, for by the synthesis gait feature ratio of target person in the real-time gait feature and gait feature library
It is right, after comparison passes through, according to the video data that target person in the video data acts, obtains and enter predeterminable area target
The real-time action feature of personnel；Wherein, the comprehensive gait feature in the gait feature library is in history video data
Each secondary real-time gait feature of target person carries out what deep learning obtained；
Enquiry module, for by the combined action aspect ratio pair of target person in the real-time action feature and motion characteristic library,
After comparison passes through, query actions feature database obtains the corresponding signature of the combined action feature, wherein motion characteristic
The corresponding relationship of combined action feature and signature is stored in library；Combined action is characterized in going through in the motion characteristic library
Each secondary real-time action feature of target person carries out what deep learning obtained in history video data；
Interactive module is interacted for obtaining machine instruction corresponding with the signature according to the machine instruction.
12. a kind of computer equipment, including memory, processor and it is stored on the memory and can be in the processor
The computer program of upper operation, which is characterized in that the processor realized when executing the computer program as claim 1 to
Exchange method described in 10 any one.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711424281.3A CN108121446B (en) | 2017-12-25 | 2017-12-25 | Exchange method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711424281.3A CN108121446B (en) | 2017-12-25 | 2017-12-25 | Exchange method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108121446A CN108121446A (en) | 2018-06-05 |
| CN108121446B true CN108121446B (en) | 2018-11-30 |
Family
ID=62231502
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711424281.3A Active CN108121446B (en) | 2017-12-25 | 2017-12-25 | Exchange method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108121446B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108921106B (en) * | 2018-07-06 | 2021-07-06 | 重庆大学 | Capsule-based face recognition method |
| CN110990523A (en) * | 2018-09-30 | 2020-04-10 | 北京国双科技有限公司 | Legal document determining method and system |
| CN113297954A (en) * | 2021-05-21 | 2021-08-24 | 中铁第四勘察设计院集团有限公司 | Ticket checking method and system for tramcar, computer equipment and readable medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101251894A (en) * | 2008-01-28 | 2008-08-27 | 天津大学 | Gait recognizing method and gait feature abstracting method based on infrared thermal imaging |
| CN103035115A (en) * | 2012-12-10 | 2013-04-10 | Tcl通力电子（惠州）有限公司 | System and method for controlling household equipment based on intelligent terminal gesture recognition |
| CN105677031A (en) * | 2016-01-04 | 2016-06-15 | 广州华欣电子科技有限公司 | Control method and device based on gesture track recognition |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111880650A (en) * | 2015-04-30 | 2020-11-03 | 谷歌有限责任公司 | Gesture recognition based on wide field radar |
-
2017
- 2017-12-25 CN CN201711424281.3A patent/CN108121446B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101251894A (en) * | 2008-01-28 | 2008-08-27 | 天津大学 | Gait recognizing method and gait feature abstracting method based on infrared thermal imaging |
| CN103035115A (en) * | 2012-12-10 | 2013-04-10 | Tcl通力电子（惠州）有限公司 | System and method for controlling household equipment based on intelligent terminal gesture recognition |
| CN105677031A (en) * | 2016-01-04 | 2016-06-15 | 广州华欣电子科技有限公司 | Control method and device based on gesture track recognition |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108121446A (en) | 2018-06-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10991236B2 (en) | Detecting of patterns of activity based on identified presence detection | |
| US11921477B2 (en) | Smart-home automation system that suggests or automatically implements selected household policies based on sensed observations | |
| US10764735B2 (en) | Methods and apparatus for using smart environment devices via application program interfaces | |
| US20210160326A1 (en) | Utilizing context information of environment component regions for event/activity prediction | |
| US10535349B2 (en) | Controlling connected devices using a relationship graph | |
| US20200167834A1 (en) | Intelligent identification and provisioning of devices and services for a smart home environment | |
| JP6858260B2 (en) | Interactive environment controller | |
| CN108121446B (en) | Exchange method and system | |
| EP3158714A1 (en) | Methods and apparatus for using smart environment devices via application program interfaces | |
| Novák et al. | Anomaly detection in user daily patterns in smart-home environment | |
| CN107431649B (en) | The method and system realized for resident's strategy in intelligent household's environment | |
| US9613514B2 (en) | Systems and methods for providing a smart notifications system | |
| CN105182772A (en) | Household intelligent service method based on wireless sensor | |
| Pazhoumand-Dar et al. | Detecting deviations from activities of daily living routines using kinect depth maps and power consumption data | |
| KR101955939B1 (en) | Method and apparatus for recognizing activities through spatial filtering and machine learning | |
| STROPPOLO | Bridge 2.0: a framework for elderly independence and ambient assisted living | |
| Dominici | Contributing to energy efficiency through a user-centered smart 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 |