CN110414393A - A kind of natural interactive method and terminal based on deep learning - Google Patents
A kind of natural interactive method and terminal based on deep learning Download PDFInfo
- Publication number
- CN110414393A CN110414393A CN201910636847.1A CN201910636847A CN110414393A CN 110414393 A CN110414393 A CN 110414393A CN 201910636847 A CN201910636847 A CN 201910636847A CN 110414393 A CN110414393 A CN 110414393A
- Authority
- CN
- China
- Prior art keywords
- click
- deep learning
- area
- clicking
- key point
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
- G06V10/235—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition based on user input or interaction
-
- 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
Abstract
The present invention provides a kind of natural interactive method and terminal based on deep learning, using deep learning model in the key point for clicking implementation body comprising clicking identification in the corresponding image of click on area for implementing body；The movement for clicking implementation body is determined according to the key point；It calculates the click and implements the region where the movement of body and the depth difference of the click on area, determine that the click implements whether body clicks the click on area according to the depth difference；It does not need clicking the default feature texture of implementation body setting, it is able to detect the click without feature and implements whether body has carried out physics click, achieve the effect that natural interaction, be very suitable to the terminal of the human-computer interactions such as point reader, game machine, improves the convenience of human-computer interaction.
Description
Technical field
The present invention relates to field of human-computer interaction more particularly to a kind of natural interactive methods and terminal based on deep learning.
Background technique
Existing point reads terminal and generallys use talking pen perhaps finger point read mode realization point reading talking pen or a finger point
Reading have the characteristics that one it is common, be exactly that talking pen or finger need to have pre-set feature texture.It is logical that point, which reads terminal,
It crosses and identifies the feature texture on talking pen or finger to determine talking pen or the specific signified position of finger and judge talking pen
Or whether finger contacts with its signified position, and carries out corresponding response according to this.
However, above-mentioned point read mode carry out will ensuring that talking pen or finger have before reading every time it is pre-set
Textural characteristics, and to guarantee the integrality of textural characteristics, otherwise, it will lead to a read error, inevitably cause one to operator
Fixed inconvenience.
Summary of the invention
The technical problems to be solved by the present invention are: a kind of natural interactive method and terminal based on deep learning is provided,
A reading can be realized without presetting feature texture, improve the convenience of human-computer interaction.
In order to solve the above-mentioned technical problem, a kind of technical solution that the present invention uses are as follows:
A kind of natural interactive method based on deep learning, comprising steps of
S1, implemented using deep learning model comprising clicking identification in the corresponding image of click on area for implementing body and clicking
The key point of body；
S2, the movement for clicking implementation body is determined according to the key point；
The depth difference in region and the click on area where S3, the calculating movement for clicking implementation body, according to described
Depth difference determines that the click implements whether body clicks the click on area.
Further, include: before the step S1
Implement body using the click for being labeled with key point information to be trained the deep learning model, makes the depth
Learning model can position the key point for clicking implementation body.
Further, the step S2 includes:
The movement for clicking implementation body is determined according to the key point using the classifier of deep learning model.
Further, the step S2 further include:
Judge that described click implements whether the movement of body is click action, if so, thening follow the steps S3.
Further, the step S3 includes:
Corresponding depth map is converted by described image by depth convolutional neural networks, obtains depth information；
The region where the movement for clicking implementation body and the depth of the click on area are calculated according to the depth information
It is poor to spend；
Determine that the click implements whether body contacts with the click on area according to the depth difference, if so, the point
It hits and implements the body click click on area, otherwise, the click implements body and do not click on the click on area.
Further, after the step S3 further include:
If body is implemented in the click clicks the click on area, described click is implemented into body to the point of the click on area
It hits and is converted into identifiable click action mode.
In order to solve the above-mentioned technical problem, the another technical solution that the present invention uses are as follows:
A kind of natural interaction terminal based on deep learning, including memory, processor and storage are on a memory and can
The computer program run on the processor, the processor perform the steps of when executing the computer program
S1, implemented using deep learning model comprising clicking identification in the corresponding image of click on area for implementing body and clicking
The key point of body；
S2, the movement for clicking implementation body is determined according to the key point；
The depth difference in region and the click on area where S3, the calculating movement for clicking implementation body, according to described
Depth difference determines that the click implements whether body clicks the click on area.
Further, include: before the step S1
Implement body using the click for being labeled with key point information to be trained the deep learning model, makes the depth
Learning model can position the key point for clicking implementation body.
Further, the step S2 includes:
The movement for clicking implementation body is determined according to the key point using the classifier of deep learning model.
Further, the step S2 further include:
Judge that described click implements whether the movement of body is click action, if so, thening follow the steps S3.
Further, the step S3 includes:
Corresponding depth map is converted by described image by depth convolutional neural networks, obtains depth information；
The region where the movement for clicking implementation body and the depth of the click on area are calculated according to the depth information
It is poor to spend；
Determine that the click implements whether body contacts with the click on area according to the depth difference, if so, the point
It hits and implements the body click click on area, otherwise, the click implements body and do not click on the click on area.
Further, after the step S3 further include:
If body is implemented in the click clicks the click on area, described click is implemented into body to the point of the click on area
It hits and is converted into identifiable click action mode.
The beneficial effects of the present invention are: the key point for implementing body, and root are clicked based on deep learning model automatic identification
It is determined according to the key point that identifies and clicks the movement for implementing body, implemented the movement of body based on described click and pass through not same district in image
Depth difference between domain, which determines to click, implements whether body clicks the click on area, does not need clicking the default spy of implementation body setting
Texture is levied, the click without feature is able to detect and implements whether body has carried out physics click, achieve the effect that natural interaction, it is non-
The terminal of the often human-computer interactions such as suitable point reader, game machine, improves the convenience of human-computer interaction.
Detailed description of the invention
Fig. 1 is a kind of step flow chart of natural interactive method based on deep learning of the embodiment of the present invention；
Fig. 2 is a kind of structural schematic diagram of natural interaction terminal based on deep learning of the embodiment of the present invention；
Fig. 3 is the schematic diagram of the training data for training deep learning model of the embodiment of the present invention；
Fig. 4 is that the embodiment of the present invention successfully detects finger and the written schematic diagram whether contacted；
Fig. 5 is that the embodiment of the present invention can not determine finger and the written schematic diagram whether contacted；
Fig. 6 is the schematic diagram of the relative depth information in image of the embodiment of the present invention between different pixels；
Label declaration:
1, a kind of natural interaction terminal based on deep learning；2, memory；3, processor.
Specific embodiment
To explain the technical content, the achieved purpose and the effect of the present invention in detail, below in conjunction with embodiment and cooperate attached
Figure is explained.
Natural interactive method and terminal proposed by the present invention based on deep learning can be suitable for any required realization people
The scene of machine interaction, such as point reader, game machine, dummy keyboard, virtual mouse, interaction, VR scene, AR scene, MR
Scape etc. illustrates below with reference to specific application scenarios.
Please refer to Fig. 1, a kind of natural interactive method based on deep learning, comprising steps of
S1, implemented using deep learning model comprising clicking identification in the corresponding image of click on area for implementing body and clicking
The key point of body；
Wherein, it is shot by a camera to comprising the click on area for clicking implementation body, obtains corresponding image, so
After acquire described image, the image of acquisition can be post-processed by ISP module, specifically, image 4A effect can be carried out
Fruit processing is clicked and implements body and can be arbitrarily to can be realized the material object of click, such as pen, hand, game paddle etc., and click on area
The interface of human-computer interaction, such as books, interactive game interface, metope, desktop etc. can be that by；
A candidate can be intercepted out in the drawings with the dynamic detection region to be detected clicked where implementing body
Then regional frame carries out identifying processing for the image in the frame of candidate region, click the key point for implementing body to identify；
Wherein, in order to intercept out candidate region, the angle point of continuum is first found, all angle points can be surrounded by then calculating
Matrix, this matrix is candidate region frame；
Deep learning model is the spy using multilayer convolution study to training data (the click picture of labeled characteristic point)
Sign；Then it is identified using data characteristics and (corresponds to the present invention, i.e. click recognition)；
Single layer network is the weight matrix of a N*N, and multitiered network is exactly M layers of weight matrix, wherein each layer of matrix
It is not equal big；
Training process is exactly by the weighed value adjusting of weight matrix N*N*M to locally or globally optimal；
Multitiered network may learn the low-level features, mid-level features and advanced features of data；
After the completion of feature identification, by classifier, i.e., weight is comprehensive can determine whether it is click action；
It, to clicking before the key point for implementing body identifies, is needed using being labeled with key using deep learning model
The click of point information is implemented body and is trained to the deep learning model, enables the deep learning model to the click
Implement body key point positioned, be illustrated in figure 3 the training data for training deep learning model comprising
As clicking the various postures and key point information on hand for implementing the hand of body；
Specifically, deep learning model includes convolutional neural networks model, convolutional neural networks can be dynamic for clicking
The key point information location tasks of work are using supervision, the semi-supervised or modes convolutional neural networks that training is completed in advance such as unsupervised
Model, trained concrete mode the present embodiment do not limit:
For example, monitor mode can be used in convolutional neural networks model, training is completed in advance, such as uses the mark of click action
The preparatory training convolutional neural networks model of data；
The network structure of convolutional neural networks model can need flexible design according to click action Information locating task,
The present embodiment is not intended to limit: for example, convolutional neural networks model can include but is not limited to convolutional layer, elu layers of linear R, pond
Change layer, full articulamentum etc., the network number of plies is more, then network is deeper；For another example, deep learning prototype network structure can use but not
It is limited to Mobilenet, depth residual error network (Deep Residual Network, ResNet) or VGGnet (Visual
Geometry Group Network) etc. networks structure；
S2, the movement for clicking implementation body is determined according to the key point；
It can recognize that according to the key point identified using the classifier of deep learning model and click the movement for implementing body,
For example, the specific posture sold can determine according to the key point recognized if clicking implementation body is hand, and herein, classification
Device can be selected according to actual needs, be also possible to other classifiers, such as Sigmoid, softmax, weighted etc.；
The depth difference in region and the click on area where S3, the calculating movement for clicking implementation body, according to described
Depth difference determines that the click implements whether body clicks the click on area；
Due to only one video camera, so to convert its correspondence for picture by specific depth convolutional neural networks
Depth map implement whether body connects with plane to be measured, such as desktop, paper etc. to judge to click to obtain depth information；
Specifically, the relative depth information among original image can be obtained by the training of a neural network, i.e., in picture
Each pixel between opposite distant relationships, as shown in Figure 6；And when the distance between different objects in picture are close
It waits, their depth information can also approach, and it is similar to will appear as color in figure, closer, and color just approaches identical, and work as it
Distance farther out when, colour-difference is away from larger, accordingly it may determine that whether different objects have contact；
In another alternative embodiment, further include in step s 2 judge it is described click implement body movement whether be
Click action, if so, thening follow the steps S3；
Implement whether the movement of body is the movement that can generate clicking operation that is, first judging to click, if so,
Just execute step S3, progress further judge click implementation body whether with plane contact to be measured, as Fig. 4,5 be respectively can
It is correct to detect and correctly detect to click the schematic diagram whether implementation body contacts with plane to be measured, so by first carrying out
The judgement for implementing body posture is clicked, whether can generate clicking operation, if not clicking operation can be generated if can prejudge in advance
Click action can not then execute the subsequent judgement whether contacted, not only increase recognition efficiency, also reduce resource consumption；
In another alternative embodiment, the step S3 includes:
Corresponding depth map is converted by described image by depth convolutional neural networks, obtains depth information；
The region where the movement for clicking implementation body and the depth of the click on area are calculated according to the depth information
It is poor to spend；
Determine that the click implements whether body contacts with the click on area according to the depth difference, if so, the point
It hits and implements the body click click on area, otherwise, the click implements body and do not click on the click on area.
In another alternative embodiment, after judging that click implementation body contacts with click on area, needing will be described
Physics is clicked to be associated with the particular content of click on area (for example interact books, interactive game etc.), so that corresponding man-machine
Interactive device can make consistent correlation behavior, for example point reads sounding, and game machine carries out corresponding response etc. after click；
Specifically, the click is implemented body to the click if the click implements body and clicks the click on area
The click in region is converted into identifiable click action mode, and identifiable click action mode refers to external system herein,
Such as point reader, game machine, the correspondence equipment of dummy keyboard connection, the point that the correspondence equipment etc. of virtual mouse connection can identify
Action mode is hit, the external system is enabled to make corresponding response operation according to the click action mode recognized, than
Such as point reader, if it is determined that there is some English word touched in written to finger, then by finger and it is written in
The contact of some English word, which is converted into, reads the English word, then after point reader recognizes this instruction, can read pair
In English word, for another example, in dummy keyboard, when recognize finger touch some letter when, then by finger and some word
Female contact is converted into the output letter, then when the equipment connecting with dummy keyboard recognizes this instruction, then output pair
The letter answered；
It is converted into the click action mode that external system can identify to physical points blow mode by above-mentioned, improves
The versatility of natural interactive method out, can be suitable for the various application scenarios for needing to carry out human-computer interaction.
Referring to figure 2., a kind of natural interaction terminal 1 based on deep learning, including memory 2, processor 3 and be stored in
On memory 2 and the computer program that can be run on the processor 3, when the processor 3 executes the computer program
The corresponding operating procedure of above-mentioned each embodiment of the method is realized respectively.
In conclusion a kind of natural interactive method and terminal based on deep learning provided by the invention, is based on depth
It practises model automatic identification and clicks the key point for implementing body, and determined according to the key point identified and click the movement for implementing body, In
It determines to click the movement for implementing body to pass through different zones in image based on the movement for clicking implementation body after click action
Between depth difference determine to click and implement whether body clicks the click on area, and the physical points blow mode is converted to and can be known
Other click action mode, so that human-computer interaction device such as point reader, interactive game machine etc. can identify the click action
Mode simultaneously makes corresponding response, does not need clicking the default feature texture of implementation body setting, can easily and efficiently detect not
Click with feature implements whether body has carried out physics click, and human-computer interaction device is made according to physics click
Corresponding response, achievees the effect that natural interaction, is very suitable to the terminal of the human-computer interactions such as point reader, game machine, improves people
The convenience of machine interaction.
The above description is only an embodiment of the present invention, is not intended to limit the scope of the invention, all to utilize this hair
Equivalents made by bright specification and accompanying drawing content are applied directly or indirectly in relevant technical field, similarly include
In scope of patent protection of the invention.
Claims (12)
1. a kind of natural interactive method based on deep learning, which is characterized in that comprising steps of
S1, body is being implemented comprising clicking identification in the corresponding image of click on area for implementing body and clicking using deep learning model
Key point；
S2, the movement for clicking implementation body is determined according to the key point；
The depth difference in region and the click on area where S3, the calculating movement for clicking implementation body, according to the depth
Difference determines that the click implements whether body clicks the click on area.
2. a kind of natural interactive method based on deep learning according to claim 1, which is characterized in that its feature exists
In the step S1 includes: before
Implement body using the click for being labeled with key point information to be trained the deep learning model, makes the deep learning
Model can position the key point for clicking implementation body.
3. a kind of natural interactive method based on deep learning according to claim 1, which is characterized in that the step S2
Include:
The movement for clicking implementation body is determined according to the key point using the classifier of deep learning model.
4. a kind of natural interactive method based on deep learning according to claim 1 or 3, which is characterized in that the step
Rapid S2 further include:
Judge that described click implements whether the movement of body is click action, if so, thening follow the steps S3.
5. a kind of natural interactive method based on deep learning according to claim 1, which is characterized in that the step S3
Include:
Corresponding depth map is converted by described image by depth convolutional neural networks, obtains depth information；
The region where the movement for clicking implementation body and the depth difference of the click on area are calculated according to the depth information；
Determine that the click implements whether body contacts with the click on area according to the depth difference, if so, clicks reality
Donor clicks the click on area, and otherwise, the click implements body and do not click on the click on area.
6. a kind of natural interactive method based on deep learning according to claim 1 or 5, which is characterized in that the step
After rapid S3 further include:
If the click implements body and clicks the click on area, by the click turn clicked and implement body to the click on area
Turn to identifiable click action mode.
7. a kind of natural interaction terminal based on deep learning, including memory, processor and storage are on a memory and can be
The computer program run on the processor, which is characterized in that the processor realized when executing the computer program with
Lower step:
S1, body is being implemented comprising clicking identification in the corresponding image of click on area for implementing body and clicking using deep learning model
Key point；
S2, the movement for clicking implementation body is determined according to the key point；
The depth difference in region and the click on area where S3, the calculating movement for clicking implementation body, according to the depth
Difference determines that the click implements whether body clicks the click on area.
8. a kind of natural interactive method based on deep learning according to claim 7, which is characterized in that its feature exists
In the step S1 includes: before
Implement body using the click for being labeled with key point information to be trained the deep learning model, makes the deep learning
Model can position the key point for clicking implementation body.
9. a kind of natural interactive method based on deep learning according to claim 7, which is characterized in that the step S2
Include:
The movement for clicking implementation body is determined according to the key point using the classifier of deep learning model.
10. a kind of natural interactive method based on deep learning according to claim 7 or 9, which is characterized in that the step
Rapid S2 further include:
Judge that described click implements whether the movement of body is click action, if so, thening follow the steps S3.
11. a kind of natural interactive method based on deep learning according to claim 7, which is characterized in that the step
S3 includes:
Corresponding depth map is converted by described image by depth convolutional neural networks, obtains depth information；
The region where the movement for clicking implementation body and the depth difference of the click on area are calculated according to the depth information；
Determine that the click implements whether body contacts with the click on area according to the depth difference, if so, clicks reality
Donor clicks the click on area, and otherwise, the click implements body and do not click on the click on area.
12. a kind of natural interactive method based on deep learning according to claim 7 or 11, which is characterized in that described
After step S3 further include:
If the click implements body and clicks the click on area, by the click turn clicked and implement body to the click on area
Turn to identifiable click action mode.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910636847.1A CN110414393A (en) | 2019-07-15 | 2019-07-15 | A kind of natural interactive method and terminal based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910636847.1A CN110414393A (en) | 2019-07-15 | 2019-07-15 | A kind of natural interactive method and terminal based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110414393A true CN110414393A (en) | 2019-11-05 |
Family
ID=68361428
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910636847.1A Pending CN110414393A (en) | 2019-07-15 | 2019-07-15 | A kind of natural interactive method and terminal based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110414393A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111027533A (en) * | 2019-12-12 | 2020-04-17 | 广东小天才科技有限公司 | Conversion method and system of point-to-read coordinates, terminal device and storage medium |
| CN114115693A (en) * | 2020-08-26 | 2022-03-01 | 华为技术有限公司 | Method and device for inputting text based on virtual keyboard |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150104075A1 (en) * | 2013-10-16 | 2015-04-16 | Qualcomm Incorporated | Z-axis determination in a 2d gesture system |
| CN104571482A (en) * | 2013-10-22 | 2015-04-29 | 中国传媒大学 | Digital device control method based on somatosensory recognition |
| CN104750286A (en) * | 2013-12-26 | 2015-07-01 | 联想(北京)有限公司 | Data acquisition method and electronic device |
| CN105787485A (en) * | 2014-12-25 | 2016-07-20 | 联想(北京)有限公司 | Identification clicking operation device and identification clicking operation method |
| CN106547356A (en) * | 2016-11-17 | 2017-03-29 | 科大讯飞股份有限公司 | Intelligent interactive method and device |
| CN107357414A (en) * | 2016-05-09 | 2017-11-17 | 株式会社理光 | A kind of recognition methods of click action and click action identification device |
| CN108227912A (en) * | 2017-11-30 | 2018-06-29 | 北京市商汤科技开发有限公司 | Apparatus control method and device, electronic equipment, computer storage media |
| CN108389226A (en) * | 2018-02-12 | 2018-08-10 | 北京工业大学 | A kind of unsupervised depth prediction approach based on convolutional neural networks and binocular parallax |
| CN108399367A (en) * | 2018-01-31 | 2018-08-14 | 深圳市阿西莫夫科技有限公司 | Hand motion recognition method, apparatus, computer equipment and readable storage medium storing program for executing |
| CN108921129A (en) * | 2018-07-20 | 2018-11-30 | 网易（杭州）网络有限公司 | Image processing method, system, medium and electronic equipment |
| CN109725724A (en) * | 2018-12-29 | 2019-05-07 | 百度在线网络技术（北京）有限公司 | There are the gestural control method and device of screen equipment |
-
2019
- 2019-07-15 CN CN201910636847.1A patent/CN110414393A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150104075A1 (en) * | 2013-10-16 | 2015-04-16 | Qualcomm Incorporated | Z-axis determination in a 2d gesture system |
| CN104571482A (en) * | 2013-10-22 | 2015-04-29 | 中国传媒大学 | Digital device control method based on somatosensory recognition |
| CN104750286A (en) * | 2013-12-26 | 2015-07-01 | 联想(北京)有限公司 | Data acquisition method and electronic device |
| CN105787485A (en) * | 2014-12-25 | 2016-07-20 | 联想(北京)有限公司 | Identification clicking operation device and identification clicking operation method |
| CN107357414A (en) * | 2016-05-09 | 2017-11-17 | 株式会社理光 | A kind of recognition methods of click action and click action identification device |
| CN106547356A (en) * | 2016-11-17 | 2017-03-29 | 科大讯飞股份有限公司 | Intelligent interactive method and device |
| CN108227912A (en) * | 2017-11-30 | 2018-06-29 | 北京市商汤科技开发有限公司 | Apparatus control method and device, electronic equipment, computer storage media |
| CN108399367A (en) * | 2018-01-31 | 2018-08-14 | 深圳市阿西莫夫科技有限公司 | Hand motion recognition method, apparatus, computer equipment and readable storage medium storing program for executing |
| CN108389226A (en) * | 2018-02-12 | 2018-08-10 | 北京工业大学 | A kind of unsupervised depth prediction approach based on convolutional neural networks and binocular parallax |
| CN108921129A (en) * | 2018-07-20 | 2018-11-30 | 网易（杭州）网络有限公司 | Image processing method, system, medium and electronic equipment |
| CN109725724A (en) * | 2018-12-29 | 2019-05-07 | 百度在线网络技术（北京）有限公司 | There are the gestural control method and device of screen equipment |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111027533A (en) * | 2019-12-12 | 2020-04-17 | 广东小天才科技有限公司 | Conversion method and system of point-to-read coordinates, terminal device and storage medium |
| CN111027533B (en) * | 2019-12-12 | 2024-02-23 | 广东小天才科技有限公司 | Click-to-read coordinate transformation method, system, terminal equipment and storage medium |
| CN114115693A (en) * | 2020-08-26 | 2022-03-01 | 华为技术有限公司 | Method and device for inputting text based on virtual keyboard |
| WO2022042449A1 (en) * | 2020-08-26 | 2022-03-03 | 华为技术有限公司 | Method and apparatus for inputting text on the basis of virtual keyboard |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Xu | A real-time hand gesture recognition and human-computer interaction system | |
| CN111563502B (en) | Image text recognition method and device, electronic equipment and computer storage medium | |
| CN108351828A (en) | Technology for device-independent automatic application test | |
| Nai et al. | Fast hand posture classification using depth features extracted from random line segments | |
| Dhawan et al. | Implementation of hand detection based techniques for human computer interaction | |
| Conseil et al. | Comparison of Fourier descriptors and Hu moments for hand posture recognition | |
| Nair et al. | Hand gesture recognition system for physically challenged people using IOT | |
| CN109919077B (en) | Gesture recognition method, device, medium and computing equipment | |
| CN111680594A (en) | Augmented reality interaction method based on gesture recognition | |
| CN111078552A (en) | Method and device for detecting page display abnormity and storage medium | |
| CN112001394A (en) | Dictation interaction method, system and device based on AI vision | |
| CN113762269A (en) | Chinese character OCR recognition method, system, medium and application based on neural network | |
| CN111399638A (en) | Blind computer and intelligent mobile phone auxiliary control method adapted to same | |
| CN112257665A (en) | Image content recognition method, image recognition model training method, and medium | |
| CN110414393A (en) | A kind of natural interactive method and terminal based on deep learning | |
| WO2023165616A1 (en) | Method and system for detecting concealed backdoor of image model, storage medium, and terminal | |
| US11157765B2 (en) | Method and system for determining physical characteristics of objects | |
| Baumgartl et al. | Vision-based hand gesture recognition for human-computer interaction using MobileNetV2 | |
| CN108628455B (en) | Virtual sand painting drawing method based on touch screen gesture recognition | |
| CN108108648A (en) | A kind of new gesture recognition system device and method | |
| WO2023102723A1 (en) | Image processing method and system | |
| CN113158870B (en) | Antagonistic training method, system and medium of 2D multi-person gesture estimation network | |
| Dhamanskar et al. | Human computer interaction using hand gestures and voice | |
| CN115223239A (en) | Gesture recognition method and system, computer equipment and readable storage medium | |
| Jeong et al. | Hand gesture user interface for transforming objects in 3d virtual space |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191105 |