CN107451559A - Parkinson's people's handwriting automatic identifying method based on machine learning - Google Patents
Parkinson's people's handwriting automatic identifying method based on machine learning Download PDFInfo
- Publication number
- CN107451559A CN107451559A CN201710640756.6A CN201710640756A CN107451559A CN 107451559 A CN107451559 A CN 107451559A CN 201710640756 A CN201710640756 A CN 201710640756A CN 107451559 A CN107451559 A CN 107451559A
- Authority
- CN
- China
- Prior art keywords
- image
- input
- iconic
- iconic model
- accuracy
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/40—Document-oriented image-based pattern recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
- G06F16/5846—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using extracted text
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Library & Information Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of Parkinson's people's handwriting automatic identifying method based on machine learning, including:1st, the hand-written word of Parkinson's people is collected in advance;2nd, by each word production into corresponding image, and unified picture specification is converted the image into;3rd, all images are classified one by one according to the content of image;4th, the image classified is established into iconic model one by one;5th, the image in recursive call iconic model is iterated training to reduce error existing for iconic model, the accuracy of each iconic model of final optimization pass;6th, when user's input picture, the image of input and the image similarity to be prestored in all iconic models is contrasted, judges whether its similarity is higher than default similarity threshold, if so, then automatically identifying the picture material of input;If it is not, word corresponding to the image of user's handwriting input input is then reminded, into step 2.The present invention can automatically identify the word content that patient Parkinson writes, and facilitate daily life.
Description
Technical field
The present invention relates to a kind of recognition methods, more particularly to a kind of Parkinson's people handwriting based on machine learning is certainly
Dynamic recognition methods.
Background technology
Parkinson's (Parkinson ' s Disease, abbreviation PD), also known as paralysis agitans syndrome.Parkinson integrates
Sign, is a kind of chronic central nervous system degenerative disorder disease.The current whole world there are about 5,000,000 Parkinson's sufferers according to statistics
Person, as Chinese aging problem is constantly aggravated, for city over-65s crowd Parkinson illness rate close to 2%, pa is golden at present
Gloomy sick serious threat the health of China's mid-aged population.Parkinsonian generally has different degrees of dyskinesia disease
Shape, clinically using static tremor, bradykinesia, splinting and posture gait disorder as principal character.Patient's Parkinson is quiet
Breath property tremble be due to the multipair Opposing muscle of upper limbs alternating contractions motion caused by, frequency of trembling typically between 3Hz-9Hz,
Signal source of trembling is located at corticocerebral neuron activity network.The generation that upper limbs tranquillization trembles, not only influence the shape of patient
Body posture, but also activity of daily living and the larger negative interaction of life quality generation to patient, such as Parkinsonian's hand
Tremble and will result in the ability decline of handwriting, influence the identification of handwriting.
For the handwriting identification technology of patient Parkinson, there has been no more intact solution at present.Traditional
Solution is mostly the installation vibration abatement on hand in patient Parkinson, is trembled with mitigating Parkinsonian's hand, and lead to
Artificial identification method is crossed to identify the handwriting of patient Parkinson.Corresponding vibration abatement prices are rather stiff, it is necessary to
Body carries, and is inconvenient, and in daily operation or on, there is inconvenience and uncertain factor.
The content of the invention
The technical problem to be solved in the present invention, it is to provide a kind of Parkinson's people handwriting based on machine learning certainly
Dynamic recognition methods, it is not necessary to install vibration abatement additional, the word content of patient's Parkinson write-in can be automatically identified, it is greatly convenient
The daily life of patient Parkinson.
What the present invention was realized in:
A kind of Parkinson's people's handwriting automatic identifying method based on machine learning, comprises the following steps:
Step 1, the hand-written word of substantial amounts of Parkinson's people is collected in advance;
Step 2, by each word production into corresponding image, and all images are converted to unified picture specification;
Step 3, all images are classified one by one according to the content of image;
Step 4, the image classified established into iconic model one by one;
Image in step 5, recursive call iconic model is iterated training to reduce error existing for iconic model, most
Optimize the accuracy of each iconic model eventually;
Step 6, when user's input picture, the image for contrasting input is similar to the image to be prestored in all iconic models
Degree, judges whether its similarity is higher than default similarity threshold, if so, then image most like in output image model and knowledge
Other result, the image for automatically identifying input are consistent with most like picture material;If it is not, then remind user handwriting input that this is defeated
Word corresponding to the image entered, into step 2.
Further, the step 4 is specially:
The image classified is converted into corresponding data, passes through the placeholder of floating number corresponding to establishment and size of data
To describe the input form of the data, and according to the input form by the data input into artificial intelligence engine, use machine
The method of device study, and after carrying out computing automatically to the data by artificial intelligence engine, iconic model corresponding to foundation,
Randomly select a plurality of key points in each iconic model on image, each pass of each image in same iconic model
Key point is corresponded, and the position of key point on each image is stored in corresponding iconic model.
Further, the step 5 is specially:
Default accuracy threshold, standard picture is defaulted as by the randomly selected in the iconic model the 1st image, is called
Image in iconic model is iterated training, by the n-th image randomly selected respectively with the 1st to (n-1)th image
Successfully any one image is identified for middle identification, reads n-th image and corresponds to key point with the image being identified therewith
Position, the position of each key point is compared, judges whether its accuracy is higher than default accuracy threshold, if identification
The accuracy arrived is higher than default accuracy threshold, then n-th image identifies successfully, is standard picture, and be stored in the image
In model;If the accuracy recognized is not higher than default accuracy threshold, n-th image is extracted from the iconic model
Out, using the automatic differentiation technique of artificial intelligence engine, the corresponding parameter of adjust automatically n-th image, and rejoin
Identification next time is carried out in the iconic model, until the accuracy of each image is higher than default essence in the iconic model
Exactness threshold value;When the accuracy of each image in the iconic model is higher than default accuracy threshold, illustrate the image
The accuracy of model is higher than default accuracy threshold, then completes the optimization of the iconic model;Wherein, n is >=2 integer.
Further, the step 6 is specially:
Default similarity threshold, loads the iconic model established, when user's input picture, the image and image of input
The picture specification of image in model is consistent, contrasts the image of input and the image similarity to be prestored in all iconic models, sentences
Whether its similarity of breaking is higher than default similarity threshold, if similarity is higher than default similarity threshold, output image mould
Most like image and corresponding recognition result in type, and feed back to user, user automatically identify the image of input with most
Similar picture material is consistent;If similarity is not higher than default similarity threshold, output identification mistake, and feeds back to use
Family, word corresponding to the image of user's handwriting input input is reminded, into step 2.
Further, also include before the step 6:
Step 51, the accuracy for calculating all iconic models, all iconic models are entered according to the size of accuracy
Row sequence.
Further, described image specification includes image size, image color, brightness of image and picture format.
The invention has the advantages that:Of the invention to be differed greatly from traditional solution, the present invention is to collect
A large amount of Parkinson's people's handwritings, and corresponding image is fabricated to, with machine learning, along with many algorithms, complete to build
Vertical and optimization iconic model process, improve the accuracy of iconic model so that patient Parkinson need not can installed additional
Handwriting input word in the case of vibration abatement, automatic identification is carried out to the word of Parkinson's person writing, greatly reduce for
Manual identified required for identifying Parkinson's people's handwriting is spent, and is very easy to the work as usual life of patient Parkinson
It is living.
Brief description of the drawings
The present invention is further illustrated in conjunction with the embodiments with reference to the accompanying drawings.
Fig. 1 is that a kind of Parkinson's people's handwriting automatic identifying method based on machine learning of the present invention performs flow
Figure.
Embodiment
To cause the present invention to become apparent, now with a preferred embodiment, and accompanying drawing is coordinated to be described in detail below.
As shown in figure 1, a kind of Parkinson's people's handwriting automatic identifying method based on machine learning, including following step
Suddenly:
Step 1, the hand-written word of substantial amounts of Parkinson's people is collected in advance;
Step 2, by each word production into corresponding image, and all images are converted to unified picture specification,
Described image specification includes image size, image color, brightness of image and picture format;Each image is set as unified rule
Lattice, to strengthen recognition accuracy;
Step 3, all images are classified one by one according to the content of image, such as:By all " big " words collected
One kind is classified as, all " small " words collected are classified as one kind, all " more " words are classified as one kind, this part of class is by people
What work was voluntarily classified;
Step 4, the image classified established into iconic model one by one;Specially:
The image classified is converted into corresponding data, passes through the placeholder of floating number corresponding to establishment and size of data
To describe the input form of the data, (floating number is that the numeral for belonging to the number of certain particular subset in rational represents, in computer
In to some any real number of approximate representation), and according to the input form by the data input (the people into artificial intelligence engine
Work intelligent engine is the core of artificial intelligence, is the fundamental way for making computer have intelligence, and its application is throughout artificial intelligence
Every field, it is mainly using conclusion, comprehensive rather than deduction), use the method (machine learning (Machine of machine learning
Learning, ML) it is a multi-field cross discipline, it is related to probability theory, statistics, Approximation Theory, convextiry analysis, algorithm complex reason
By etc. multi-door subject.The learning behavior that the mankind were simulated or realized to computer how is specialized in, to obtain new knowledge or skills,
The existing structure of knowledge is reorganized to be allowed to constantly improve the performance of itself), and by artificial intelligence engine to the data from
After dynamic progress computing, iconic model corresponding to foundation, a plurality of key points are randomly selected on image in each iconic model, together
Each key point of each image in one iconic model is corresponded, and the position of key point on each image is stored in
In corresponding iconic model;Such as:All " big " words classified are added in the iconic model for there are respective markers, set " big "
The end points of each stroke of word and crosspoint are key point (7 points), then own " big " in the iconic model containing all " big " words
The position that these key points are all gathered on word image is stored;By all " small " classified, word is added to respective markers
In iconic model, set each stroke of " small " word end points and crosspoint as key point (6 points), then containing all " small " words
The position for all gathering these key points in iconic model on all " small " word images is stored;By all " more " words classified
Be added in the iconic model for having respective markers, set each stroke of " more " word end points and crosspoint as key point (14
Point), then deposited containing the position for all gathering these key points in the iconic model for owning " more " words on all " more " word images
Storage;
Image in step 5, recursive call iconic model is iterated training to reduce error existing for iconic model, most
Optimize the accuracy of each iconic model eventually;Specially:
Default accuracy threshold is 50%, and the randomly selected in the iconic model the 1st image is defaulted as into standard drawing
Picture, call iconic model in image be iterated training, by the n-th image randomly selected respectively with the 1st to (n-1)th
Identify that successfully any one image is identified in individual image, it is corresponding with the image being identified therewith to read n-th image
The position of key point, the position of each key point is compared, judges whether its accuracy is higher than 50%, if the essence recognized
Exactness is higher than 50%, then n-th image identifies successfully, is standard picture, and be stored in the iconic model;If the essence recognized
Exactness is not higher than 50%, then n-th image is extracted from the iconic model, use the automatic differentiation of artificial intelligence engine
Technology, the corresponding parameter of adjust automatically n-th image, and the identification carried out next time is rejoined in the iconic model, directly
The accuracy of each image is higher than 50% in the iconic model;When the accuracy of each image in the iconic model
During higher than 50%, illustrate that the accuracy of the iconic model is higher than 50%, then complete the optimization of the iconic model;Wherein, n for >=
2 integer;
Such as:The 2nd image is extracted, the 2nd image can only be identified with the 1st image, read the 1st image and the 2nd
The position of the key point of individual image, the 2nd image is compared with the position of the 1st corresponding key point of image, the essence of the two
Exactness is 60% (> 50%), then the 2nd image recognition success, is standard picture;Continue to extract the 3rd image, the 3rd image
It is identified with the 1st image or the 2nd image, selectes the 3rd image and be identified with the 2nd image, reads the 2nd figure
The position of the key point of picture and the 3rd image, the 3rd image is compared with the position of the 2nd corresponding key point of image, two
The accuracy of person is 40% (< 50%), the 3rd image is extracted from iconic model, using artificial intelligence engine from
Dynamic differentiation technique, the 3rd corresponding parameter of image of adjust automatically, and the knowledge carried out next time is reentered in the iconic model
Not;
Step 51, the accuracy for calculating all iconic models, all iconic models are entered according to the size of accuracy
Row sequence;
Step 6, when user's input picture, the image for contrasting input is similar to the image to be prestored in all iconic models
Degree, judges whether its similarity is higher than default similarity threshold, if so, then image most like in output image model and knowledge
Other result, the image for automatically identifying input are consistent with most like picture material;If it is not, then remind user handwriting input that this is defeated
Word corresponding to the image entered, into step 2;Specially:
Default similarity threshold is 50%, loads the iconic model established, when user's input picture, the image of input
Consistent with the picture specification of the image in iconic model, the image for contrasting input is similar to the image to be prestored in all iconic models
Degree, judges whether its similarity is higher than 50%, if similarity is higher than 50%, most like image and institute in output image model
Corresponding recognition result, and user is fed back to, the image that user automatically identifies input is consistent with most like picture material;If
Similarity is not higher than 50%, then output identification mistake, and feed back to user, reminds the image of user's handwriting input input corresponding
Word, into step 2.
Although the foregoing describing the embodiment of the present invention, those familiar with the art should manage
Solution, the specific embodiment described by us are merely exemplary, rather than for the restriction to the scope of the present invention, are familiar with this
The equivalent modification and change that the technical staff in field is made in the spirit according to the present invention, should all cover the present invention's
In scope of the claimed protection.
Claims (6)
- A kind of 1. Parkinson's people's handwriting automatic identifying method based on machine learning, it is characterised in that:Including following step Suddenly:Step 1, the hand-written word of substantial amounts of Parkinson's people is collected in advance;Step 2, by each word production into corresponding image, and all images are converted to unified picture specification;Step 3, all images are classified one by one according to the content of image;Step 4, the image classified established into iconic model one by one;Image in step 5, recursive call iconic model is iterated training to reduce error existing for iconic model, final excellent Change the accuracy of each iconic model;Step 6, when user's input picture, contrast the image of input and the image similarity to be prestored in all iconic models, sentence Whether its similarity of breaking is higher than default similarity threshold, if so, then image most like in output image model and identification are tied Fruit, the image for automatically identifying input are consistent with most like picture material;If it is not, then remind user's handwriting input input Word corresponding to image, into step 2.
- 2. a kind of Parkinson's people's handwriting automatic identifying method based on machine learning according to claim 1, its It is characterised by:The step 4 is specially:The image classified is converted into corresponding data, retouched by the placeholder of floating number corresponding to establishment and size of data State the input form of the data, and according to the input form by the data input into artificial intelligence engine, use engineering The method of habit, and after carrying out computing automatically to the data by artificial intelligence engine, iconic model corresponding to foundation, each Randomly select a plurality of key points in iconic model on image, each key point of each image in same iconic model Correspond, the position of key point on each image is stored in corresponding iconic model.
- 3. a kind of Parkinson's people's handwriting automatic identifying method based on machine learning according to claim 1, its It is characterised by:The step 5 is specially:Default accuracy threshold, standard picture, calling figure picture are defaulted as by the randomly selected in the iconic model the 1st image Image in model is iterated training, by the n-th image randomly selected respectively with knowing in the 1st to (n-1)th image Not successfully any one image is not identified, and reads the position that n-th image corresponds to key point with the image being identified therewith Put, the position of each key point is compared, judge whether its accuracy is higher than default accuracy threshold, if recognize Accuracy is higher than default accuracy threshold, then n-th image identifies successfully, is standard picture, and be stored in the iconic model In;If the accuracy recognized is not higher than default accuracy threshold, n-th image is extracted from the iconic model Come, using the automatic differentiation technique of artificial intelligence engine, the corresponding parameter of adjust automatically n-th image, and rejoin this Identification next time is carried out in iconic model, until the accuracy of each image is accurate higher than default in the iconic model Spend threshold value;When the accuracy of each image in the iconic model is higher than default accuracy threshold, illustrate the image mould The accuracy of type is higher than default accuracy threshold, then completes the optimization of the iconic model;Wherein, n is >=2 integer.
- 4. a kind of Parkinson's people's handwriting automatic identifying method based on machine learning according to claim 1, its It is characterised by:The step 6 is specially:Default similarity threshold, loads the iconic model established, when user's input picture, the image and iconic model of input In image picture specification it is consistent, contrast the image of input and the image similarity to be prestored in all iconic models, judge it Whether similarity is higher than default similarity threshold, if similarity is higher than default similarity threshold, in output image model Most like image and corresponding recognition result, and feed back to user, user automatically identify the image of input with it is most like Picture material it is consistent;If similarity is not higher than default similarity threshold, output identification mistake, and feeds back to user, carry Word corresponding to the image of awake user's handwriting input input, into step 2.
- 5. a kind of Parkinson's people's handwriting automatic identifying method based on machine learning according to claim 1, its It is characterised by:Also include before the step 6:Step 51, the accuracy for calculating all iconic models, all iconic models are arranged according to the size of accuracy Sequence.
- 6. a kind of Parkinson's people's handwriting automatic identifying method based on machine learning according to claim 1, its It is characterised by:Described image specification includes image size, image color, brightness of image and picture format.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710640756.6A CN107451559A (en) | 2017-07-31 | 2017-07-31 | Parkinson's people's handwriting automatic identifying method based on machine learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710640756.6A CN107451559A (en) | 2017-07-31 | 2017-07-31 | Parkinson's people's handwriting automatic identifying method based on machine learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107451559A true CN107451559A (en) | 2017-12-08 |
Family
ID=60490061
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710640756.6A Pending CN107451559A (en) | 2017-07-31 | 2017-07-31 | Parkinson's people's handwriting automatic identifying method based on machine learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107451559A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109050085A (en) * | 2018-09-25 | 2018-12-21 | 绵阳鼎飞益电子科技有限公司 | A kind of pen used for parkinsonism patient |
CN109278441A (en) * | 2018-09-25 | 2019-01-29 | 绵阳鼎飞益电子科技有限公司 | A method of for parkinsonism, patient is write |
CN109472234A (en) * | 2018-11-01 | 2019-03-15 | 北京爱知之星科技股份有限公司 | A kind of method of handwriting input intelligent recognition |
CN113743105A (en) * | 2021-09-07 | 2021-12-03 | 深圳海域信息技术有限公司 | Character similarity retrieval analysis method based on big data feature recognition |
JP7421703B2 (en) | 2020-02-21 | 2024-01-25 | 公立大学法人会津大学 | Classification program, classification device and classification method |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8472719B2 (en) * | 2002-12-17 | 2013-06-25 | Abbyy Software Ltd. | Method of stricken-out character recognition in handwritten text |
CN105335760A (en) * | 2015-11-16 | 2016-02-17 | 南京邮电大学 | Image number character recognition method |
CN106408038A (en) * | 2016-09-09 | 2017-02-15 | 华南理工大学 | Rotary Chinese character identifying method based on convolution neural network model |
CN106529525A (en) * | 2016-10-14 | 2017-03-22 | 上海新同惠自动化系统有限公司 | Chinese and Japanese handwritten character recognition method |
CN106845358A (en) * | 2016-12-26 | 2017-06-13 | 苏州大学 | A kind of method and system of handwritten character characteristics of image identification |
CN106951832A (en) * | 2017-02-28 | 2017-07-14 | 广东数相智能科技有限公司 | A kind of verification method and device based on Handwritten Digits Recognition |
-
2017
- 2017-07-31 CN CN201710640756.6A patent/CN107451559A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8472719B2 (en) * | 2002-12-17 | 2013-06-25 | Abbyy Software Ltd. | Method of stricken-out character recognition in handwritten text |
CN105335760A (en) * | 2015-11-16 | 2016-02-17 | 南京邮电大学 | Image number character recognition method |
CN106408038A (en) * | 2016-09-09 | 2017-02-15 | 华南理工大学 | Rotary Chinese character identifying method based on convolution neural network model |
CN106529525A (en) * | 2016-10-14 | 2017-03-22 | 上海新同惠自动化系统有限公司 | Chinese and Japanese handwritten character recognition method |
CN106845358A (en) * | 2016-12-26 | 2017-06-13 | 苏州大学 | A kind of method and system of handwritten character characteristics of image identification |
CN106951832A (en) * | 2017-02-28 | 2017-07-14 | 广东数相智能科技有限公司 | A kind of verification method and device based on Handwritten Digits Recognition |
Non-Patent Citations (4)
Title |
---|
周星辰: "《基于深度模型的脱机手写体汉字识别研究》", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
张炘中等: "《汉字识别的特征点法及其一种应用》", 《中文信息学报》 * |
钱国良等: "《基于机器学习的手写汉字识别的研究》", 《模式识别与人工智能》 * |
高学等: "《一种基于支持向量机的手写汉字识别方法》", 《电子学报》 * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109050085A (en) * | 2018-09-25 | 2018-12-21 | 绵阳鼎飞益电子科技有限公司 | A kind of pen used for parkinsonism patient |
CN109278441A (en) * | 2018-09-25 | 2019-01-29 | 绵阳鼎飞益电子科技有限公司 | A method of for parkinsonism, patient is write |
CN109278441B (en) * | 2018-09-25 | 2020-10-23 | 绵阳鼎飞益电子科技有限公司 | Method for writing for Parkinson's syndrome patient |
CN109472234A (en) * | 2018-11-01 | 2019-03-15 | 北京爱知之星科技股份有限公司 | A kind of method of handwriting input intelligent recognition |
CN109472234B (en) * | 2018-11-01 | 2021-07-20 | 北京爱知之星科技股份有限公司 | Intelligent recognition method for handwriting input |
JP7421703B2 (en) | 2020-02-21 | 2024-01-25 | 公立大学法人会津大学 | Classification program, classification device and classification method |
CN113743105A (en) * | 2021-09-07 | 2021-12-03 | 深圳海域信息技术有限公司 | Character similarity retrieval analysis method based on big data feature recognition |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107451559A (en) | Parkinson's people's handwriting automatic identifying method based on machine learning | |
US11967151B2 (en) | Video classification method and apparatus, model training method and apparatus, device, and storage medium | |
CN106599800A (en) | Face micro-expression recognition method based on deep learning | |
CN109597891A (en) | Text emotion analysis method based on two-way length Memory Neural Networks in short-term | |
CN109684912A (en) | A kind of video presentation method and system based on information loss function | |
CN108960080A (en) | Based on Initiative Defense image to the face identification method of attack resistance | |
Zhang et al. | An novel end-to-end network for automatic student engagement recognition | |
CN109344759A (en) | A kind of relatives' recognition methods based on angle loss neural network | |
Zeng et al. | CNN model design of gesture recognition based on tensorflow framework | |
CN109583331B (en) | Deep learning-based accurate positioning method for positions of wrist vein and mouth of person | |
CN109753571A (en) | A kind of scene map lower dimensional space embedding grammar based on secondary theme space projection | |
CN109242013A (en) | A kind of data mask method, device, electronic equipment and storage medium | |
CN110147878A (en) | Data processing method, device and equipment | |
CN109344888A (en) | A kind of image-recognizing method based on convolutional neural networks, device and equipment | |
CN109685037A (en) | A kind of real-time action recognition methods, device and electronic equipment | |
CN111582342B (en) | Image identification method, device, equipment and readable storage medium | |
CN110443372A (en) | A kind of transfer learning method and system based on entropy minimization | |
CN106650804A (en) | Facial sample cleaning method and system based on deep learning features | |
CN109325513A (en) | A kind of image classification network training method based on magnanimity list class single image | |
CN108154156A (en) | Image Ensemble classifier method and device based on neural topic model | |
CN107392251A (en) | A kind of method that target detection network performance is lifted using category images | |
CN110889335A (en) | Human skeleton double-person interaction behavior recognition method based on multi-channel space-time fusion network | |
CN114241270A (en) | Intelligent monitoring method, system and device for home care | |
Serbaya | Analyzing the role of emotional intelligence on the performance of small and medium enterprises (SMEs) using ai-based convolutional neural networks (CNNs) | |
CN111797856B (en) | Modeling method and device, storage medium and electronic equipment |
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 |
Application publication date: 20171208 |
|
RJ01 | Rejection of invention patent application after publication |