CN109656358A - A kind of multidimensional sign Language Recognition Method - Google Patents
A kind of multidimensional sign Language Recognition Method Download PDFInfo
- Publication number
- CN109656358A CN109656358A CN201811403415.8A CN201811403415A CN109656358A CN 109656358 A CN109656358 A CN 109656358A CN 201811403415 A CN201811403415 A CN 201811403415A CN 109656358 A CN109656358 A CN 109656358A
- Authority
- CN
- China
- Prior art keywords
- data
- sign language
- algorithm
- result
- group
- 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
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- 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
- G06F3/011—Arrangements for interaction with the human body, e.g. for user immersion in virtual reality
- G06F3/014—Hand-worn input/output arrangements, e.g. data gloves
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2203/00—Indexing scheme relating to G06F3/00 - G06F3/048
- G06F2203/01—Indexing scheme relating to G06F3/01
- G06F2203/011—Emotion or mood input determined on the basis of sensed human body parameters such as pulse, heart rate or beat, temperature of skin, facial expressions, iris, voice pitch, brain activity patterns
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Probability & Statistics with Applications (AREA)
- Human Computer Interaction (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of multidimensional sign Language Recognition Methods, data are acquired by using the multidimensional sensor including Wireless 3 D acceleration transducer, Wireless 3 D angular transducer, Wireless 3 D angular-rate sensor, myoelectric sensor, whole input is split using ad hoc approach, then each individual character segment is identified, and the individual character that will identify that according to word order synthesizes whole sentence, realizes and identifies to the entirety of sign language.Sound localization technology has been merged simultaneously, user can have been allowed quickly to position the position of communicatee, convenient for users to exchanging use with other people.The present invention uses Multi-sensor Fusion, and the advantage and disadvantage between complementary each sensor simplify the design of algorithm while improving algorithm discrimination, and comfort level when user uses is improved using wireless sensor, can unlimited occasion use, it is easy to operate, have a wide range of application.
Description
Technical field
The present invention relates to Sign Language Recognition technical field, especially a kind of multidimensional sign Language Recognition Method.
Background technique
Currently, the sign language exchange in the barrier human world is listened still to establish on the autonomous learning to sign language, to being unfamiliar with the outstanding of sign language
It is that abled person puts to no little inconvenience, and the study of sign language needs to expend more time and energy, and common abled person is due to answering
Less with occasion, the difficulty for popularizing sign language is larger, it is therefore desirable to a kind of sign language translation device that can be easy.
Especially the development of microelectric technique, multi-sensor fusion technology mention for hearing-impaired people with the development of science and technology
A kind of media of communication of simplicity is supplied, it is turned sign language movement by the acquisition identification to arm and the various types data of hand
It is changed to the information that abled person can be easily recognized.But current sign language interpretation system is acquired using single-sensor mostly, and only
It is simply integrally to be input to data in algorithm, discrimination is low and time-consuming, therefore does not obtain large-scale promotion.
Chinese patent 201410701079.0 discloses a kind of dress that hearing-impaired people's voice messaging is informed by vibration signal
It sets, the voice signal in ambient enviroment is acquired by acquisition module, voice messaging is converted into corresponding vibration signal after identification
Inform hearing-impaired people.Although this method can extract voice messaging, it is crucial to be only able to achieve the voice that identification is previously set
Word identification, and cannot judge sound bearing.
Summary of the invention
The purpose of the present invention is to provide a kind of multidimensional sign Language Recognition Methods of Multi-sensor Fusion.The present invention is retaining respectively
The shortcomings that compensating for different sensors while sensor advantage, and feature extraction is carried out to improve algorithm to acquisition data processing
Recognition speed.
To achieve the goals above, The technical solution adopted by the invention is as follows:
A kind of multidimensional sign Language Recognition Method, comprising the following steps:
Step 1: the creation for carrying out Sign Language Recognition template simultaneously saves the template that creation generates, as sign Language Recognition
Identification judgment basis;
Step 2: carrying out data acquisition to practical sign language, acquires three-dimensional acceleration by Wireless 3 D acceleration transducer
Data are acquired three-dimensional perspective data by Wireless 3 D angular transducer, are three-dimensional by the acquisition of Wireless 3 D angular-rate sensor
Angular velocity data and the number that totally 10 data including one-dimensional myoelectricity data-signal are one group of sign language is acquired by myoelectric sensor
According to, and collected data are inputted into sign Language Recognition;
Step 3: pre-processing collected data, including data are successively carried out with smothing filtering, normalization, is risen
Stop judgement operation;
Step 4: carrying out data type separation, the segmentation of whole sentence, individual character identification and individual character to pretreated data and integrate,
Data type separation is that pretreated data are divided into two classes: acceleration information, angle-data, angular velocity data are posture number
According to class, myoelectricity data constitute a class by itself, and then carry out whole sentence segmentation, the part of the data-intensive appearance of myoelectricity are detected, by these parts
It respectively is isolated by and, save as individual data group, these data groups are individual character group, carry out feature extraction and algorithm to individual character group
Identification, the recognition result of individual character group is stitched together in chronological order, forms a whole result, and the whole result is aobvious with text
Show, indicates character representated by sign language;
Step 5: carrying out result judgement to the whole result of text importing, modify template if result is incorrect, if knot
Fruit correctly then exports result.
Further, the data for acquiring multiple groups gesture motion in the step 1 by multidimensional sensor, by the multiple groups
The data of gesture motion input sign Language Recognition, and identifying system is by the corresponding meaning of data of judgement input to obtain itself
It is identifying as a result, the corresponding physical meaning of the data of the multiple groups gesture motion is inputted identifying system, identifying system root again
The difference of the result and physical meaning that identify according to itself is constantly corrected that itself is identified as a result, when recognition result tends towards stability
When, form Sign Language Recognition template.
Further, the Sign Language Recognition template includes identification feature library and algorithm recognition template, to collected data
It is pre-processed, including data is successively carried out with smothing filtering, normalization, start-stop point judgement operation, to pretreated data
Specific features are extracted according to feature database, using clustering algorithm by tagsort, by the physical meaning of classification results and input data
Compare, each feature will individually provide discrimination, when discrimination is in k-means clustering algorithm, k-medoids clustering algorithm, k-
Medians clustering algorithm, gauss hybrid models clustering algorithm composition four kinds of algorithms in be more than respectively in three kinds of algorithm above
Think that feature is available when 70%, keeping characteristics, institute's feature with a grain of salt forms identification feature library, and identification feature library is by acceleration degree
According to, angle-data, angular velocity data and myoelectricity data are retained separately;The feature extracted is calculated using three-layer neural network
Method identification, three-layer neural network algorithm identify the Fusion Features of data, and obtain an identification knot for one group of sign language data
Fruit, three-layer neural network algorithm modify three-layer neural network algorithm according to the deviation of recognition result and the physical meaning of input data
Parameter terminates parameter modification when final recognition correct rate is greater than 70%, saves algorithm recognition template at this time.
Further, the start-stop point in the step 3 judges operation are as follows: to the data point that 10 data are one group of sign language
Group, which is enumerated to correspond to, is divided into 10 groups of data, and starting point is judged as that a certain moment proposes the corresponding data in 10 groups of data, will be every
The value for the data that group proposes and the 20th data value before the group do difference, if there is the absolute value of any group of the difference
Greater than 0.3, then the moment is determined as starting point, after starting point, extracts the corresponding data in 10 groups of data of another moment
And the 15th data value after the value for the data that every group proposes and the group is done into difference, if the absolute value of all groups of difference
It is all larger than 0.12 and is all satisfied after data value and the data the 15th data value in 80 data after another moment
The absolute value of difference then judges the moment for end point less than 0.12.
Further, in the step 4, to individual character group carry out feature extraction include be loaded into Sign Language Recognition template after according to
Specific features are extracted in the identification feature library of Sign Language Recognition template, and algorithm identification makes according to the algorithm recognition template of Sign Language Recognition template
Algorithm identification is carried out with three-layer neural network.
Further, in the step 4, intensively occur referring to that the myoelectricity data after normalization are exhausted in the numerical value at certain time point
Available point is known as more than 0.12 to value, 10 points are referred to as to have if number is more than 5 in continuous 10 data points if available point
Data segment is imitated, if multiple valid data sections are uninterruptedly to be connected, it is referred to as close that multiple valid data section connects the data segment to be formed
Collection occurs.
Compared with prior art, the present invention its remarkable advantage is: general attitude transducer includes acceleration transducer, angle
Velocity sensor, angular transducer can accurately reflect the kinetic property of wrist, arm segment comprehensively, but can not be preferably anti-
The kinetic property of finger is reflected, and myoelectric sensor can be well reflected the kinetic property of finger part, but to arm, wrist
The shortcomings that kinetic property is insensitive, and the present invention compensates for different sensors while retaining each sensor advantage, and to acquisition
Data processing carries out feature extraction to improve algorithm recognition speed;The multidimensional sign Language Recognition Method of invention, recognition speed is high, identification
Accuracy rate, favorable expandability, adaptivity is good, meanwhile, user can also quickly more new template makes it according to oneself use habit
It is adapted to user itself use;The present invention uses Multi-sensor Fusion, and the advantage and disadvantage between complementary each sensor are improving
The design of algorithm is simplified while algorithm discrimination, and comfort level when user uses is improved using wireless sensor,
Can unlimited occasion use, it is easy to operate, have a wide range of application.
Detailed description of the invention
Fig. 1 is the flow chart of multidimensional sign Language Recognition Method of the present invention.
Fig. 2 is present invention acquisition hand ring diagram.
Specific embodiment
With reference to the accompanying drawings of the specification, the present invention is further illustrated.
In conjunction with Fig. 1, a kind of multidimensional sign Language Recognition Method, comprising the following steps:
Step 1: the creation for carrying out Sign Language Recognition template simultaneously saves the template that creation generates, as sign Language Recognition
Identification judgment basis.
Step 2: carrying out data acquisition to practical sign language, passes through Wireless 3 D acceleration by the acquisition bracelet of Fig. 2 signal
Sensor acquisition three-dimensional acceleration data acquires three-dimensional perspective data by Wireless 3 D angular transducer, passes through Wireless 3 D
Angular-rate sensor acquisition three-dimensional angular velocity data and acquired by myoelectric sensor including one-dimensional myoelectricity data-signal totally 10
A data are the data of one group of sign language, and collected data are inputted sign Language Recognition, and all sensors are mounted on Fig. 2 institute
On the front end bracelet shown;
Step 3: pre-processing collected data, including data are successively carried out with smothing filtering, normalization, is risen
Stop judgement operation.The smooth function that smothing filtering uses matlab to carry, it will reduce value data variation degree.Return
One changes the mapminmax function carried using matlab, it will making data bi-directional scaling to maximum value 1, minimum value is 0
In range.Start-stop point judgement will judge the standing part of data and reject that standing partially refers to static placement when not doing sign language
The period (time such as using the time before acquisition bracelet and after having used) of sensor;
Step 4: carrying out data type separation, the segmentation of whole sentence, individual character identification and individual character to pretreated data and integrate,
Data type separation is that pretreated data are divided into two classes: acceleration information, angle-data, angular velocity data are posture number
According to class, myoelectricity data constitute a class by itself, and then carry out whole sentence segmentation, the part of the data-intensive appearance of myoelectricity are detected, by these parts
It respectively is isolated by and, save as individual data group, these data groups are individual character group, carry out feature extraction and algorithm to individual character group
Identification, the recognition result of individual character group is stitched together in chronological order, forms a whole result, and the whole result is aobvious with text
Show, indicates character representated by sign language;
Step 5: carrying out result judgement to the whole result of text importing, modify template if result is incorrect, if knot
Fruit correctly then exports result.
Be in the step 1 acquired by multidimensional sensor on the bracelet of the front end Fig. 2 the data of multiple groups gesture motion to
The data of the multiple groups gesture motion are inputted sign Language Recognition by the acquisition for completing mass data, and identifying system will judge defeated
The corresponding meaning of the data entered is that itself is identified as a result, again by the corresponding reality of data of the multiple groups gesture motion to obtain
Border meaning inputs identifying system, and the difference of result and physical meaning that identifying system is identified according to itself is constantly corrected itself and known
Not Chu as a result, when recognition result tends towards stability, the training of template can be completed, form Sign Language Recognition template.
The Sign Language Recognition template includes identification feature library and algorithm recognition template, is located in advance to collected data
Reason, including data are successively carried out with smothing filtering, normalization, start-stop point judgement operation.To pretreated data according to feature
Extract specific features in library.The feature database be conventional statistic eigenvalue cluster at, such as average value, variance, autoregressive coefficient, from phase
Relationship number etc., the statistical characteristics in all feature databases are single output valve, and the statistical characteristics that multiple numerical value are formed
Gradually single output.Using clustering algorithm by tagsort, by classification results compared with the physical meaning of input data, Mei Gete
Sign will individually provide discrimination, when discrimination is in k-means clustering algorithm, k-medoids clustering algorithm, k-medians cluster
Algorithm, gauss hybrid models cluster (GMM) algorithm composition four kinds of algorithms in three kinds of algorithm above respectively more than 70% when
Think that feature is available, keeping characteristics, institute's feature with a grain of salt forms identification feature library, and acceleration information, angle are pressed in identification feature library
Degree evidence, angular velocity data and myoelectricity data are retained separately;Algorithm knowledge is carried out to the feature extracted using three-layer neural network
Not, three-layer neural network algorithm identifies the Fusion Features of data, and obtains a recognition result for one group of sign language data.Three
Layer neural network algorithm modifies three-layer neural network algorithm parameter according to the deviation of recognition result and the physical meaning of input data,
Terminate parameter modification when final recognition correct rate is greater than 70%, saves algorithm recognition template at this time.Finally formed sign language
Recognition template is parameter setting file, is formed in a template file.
Start-stop point in the step 3 judges operation are as follows: enumerates correspondence to the data grouping that 10 data are one group of sign language
It is divided into 10 groups of data, starting point is judged as that a certain moment proposes the corresponding data in 10 groups of data, the number that every group is proposed
According to value and the group before the 20th data value do difference, if there is any group of the difference absolute value be greater than 0.3, then
The moment is determined as starting point, after starting point, the corresponding data in 10 groups of data of another moment is extracted and mentions every group
The 15th data value after the value of data out and the group does difference, if the absolute value of all groups of difference is all larger than 0.12
And be all satisfied in 80 data after another moment data value with after the data difference of the 15th data value it is absolute
Value then judges the moment for end point less than 0.12.Since start-stop point judges to influence, the step 2 data must be when inputting
Front end bracelet is set to stand 1~2 second time relative to ground after formally doing before sign language and finishing sign language, the position where the bracelet of front end
And posture, without doing specific adjusted, posture refers to bracelet first for the angle on ground herein.
In the step 4, carrying out feature extraction to individual character group includes being loaded into after Sign Language Recognition template according to Sign Language Recognition mould
Specific features are extracted in the identification feature library of plate, are only once extracted, not by the physical meaning of classification results and input data ratio
Compared with;Algorithm, which is identified, carries out algorithm identification using three-layer neural network according to the algorithm recognition template of Sign Language Recognition template, only carries out
Primary identification, the deviation of the physical meaning of basis and input data does not modify three-layer neural network algorithm parameter.
In the step 4, intensively occur referring to that the myoelectricity data after normalization are more than in the absolute value at certain time point
0.12 is known as available point, and 10 points are referred to as valid data section if if available point, number is more than 5 in continuous 10 data points,
If multiple valid data sections are uninterruptedly to be connected, multiple valid data section connects the data segment to be formed and is known as intensively occurring.
In the step 5, modification template refers to that sign Language Recognition will use first when recognition result is not consistent with physical meaning
Preceding identification feature library, neural network will identify data in the case where being loaded into previous algorithm recognition template and change algorithm ginseng
Number.
By acoustic fix ranging module and identification of sound source module, amplitude and phase property to the specified sound-source signal received into
Row is extracted and is resolved, and obtains the location information of specified sound source;The feature of the specified sound-source signal received is analyzed and processed,
Obtain the identification information of target position.
The identification feature library is the characteristic value sequence by screening, can be by application developer, quadratic program exploit person
Member's setting, and can be updated according to recognition result.
As described above, the present invention can be better realized, above-described embodiment is only presently preferred embodiments of the present invention, is not used
To limit the scope of implementation of the present invention;It is i.e. all according to equivalent changes and modifications made by the content of present invention, be all the claims in the present invention
Range claimed is covered.
Claims (6)
1. a kind of multidimensional sign Language Recognition Method, which comprises the following steps:
Step 1: carrying out the creation of Sign Language Recognition template and saves the template that creation generates, using the knowledge as sign Language Recognition
Other judgment basis;
Step 2: to practical sign language carry out data acquisition, by Wireless 3 D acceleration transducer acquire three-dimensional acceleration data,
Three-dimensional perspective data are acquired by Wireless 3 D angular transducer, three-dimensional angular velocity is acquired by Wireless 3 D angular-rate sensor
Data and the data that totally 10 data including one-dimensional myoelectricity data-signal are one group of sign language are acquired by myoelectric sensor, and will
Collected data input sign Language Recognition;
Step 3: pre-processing collected data, including successively carries out smothing filtering, normalization, start-stop point to data
Judgement operation;
Step 4: carrying out data type separation, the segmentation of whole sentence, individual character identification and individual character to pretreated data and integrate, data
Type separation is that pretreated data are divided into two classes: acceleration information, angle-data, angular velocity data are attitude data
Class, myoelectricity data constitute a class by itself, and then carry out whole sentence segmentation, detect the part of the data-intensive appearance of myoelectricity, by these parts point
It does not separate, saves as individual data group, these data groups are individual character group, carry out feature extraction to individual character group and algorithm is known
Not, the recognition result of individual character group is stitched together in chronological order, forms a whole result, the whole result is aobvious with text
Show, indicates character representated by sign language;
Step 5: result judgement is carried out to the whole result of text importing, template is modified if result is incorrect, if result is just
It is true then export result.
2. multidimensional sign Language Recognition Method according to claim 1, which is characterized in that sensed in the step 1 by multidimensional
Device acquires the data of multiple groups gesture motion, the data of the multiple groups gesture motion is inputted sign Language Recognition, identifying system will
Judge that the corresponding meaning of data of input is that itself is identified as a result, again by the data pair of the multiple groups gesture motion to obtain
The physical meaning input identifying system answered, the difference of result and physical meaning that identifying system is identified according to itself are constantly corrected
It is that itself is identified as a result, when recognition result tends towards stability, form Sign Language Recognition template.
3. multidimensional sign Language Recognition Method according to claim 2, which is characterized in that the Sign Language Recognition template includes identification
Feature database and algorithm recognition template, pre-process collected data, including successively carry out smothing filtering, normalizing to data
Change, start-stop point judgement operation, specific features are extracted according to feature database to pretreated data, is divided feature using clustering algorithm
Class, by classification results compared with the physical meaning of input data, each feature will individually provide discrimination, when discrimination is in k-
Means clustering algorithm, k-medoids clustering algorithm, k-medians clustering algorithm, gauss hybrid models clustering algorithm composition
Four kinds of algorithms in three kinds of algorithm above respectively more than 70% when think that feature is available, keeping characteristics, institute's feature with a grain of salt
Identification feature library is formed, identification feature library is retained separately by acceleration information, angle-data, angular velocity data and myoelectricity data;
Algorithm identification is carried out to the feature extracted using three-layer neural network, three-layer neural network algorithm knows the Fusion Features of data
Not, and for one group of sign language data a recognition result is obtained, three-layer neural network algorithm is according to recognition result and input data
The deviation of physical meaning modify three-layer neural network algorithm parameter, terminate parameter when final recognition correct rate is greater than 70% and repair
Change, saves algorithm recognition template at this time.
4. multidimensional sign Language Recognition Method according to claim 3, which is characterized in that the start-stop point judgement in the step 3
Operation are as follows: correspondence is enumerated to the data grouping that 10 data are one group of sign language and is divided into 10 groups of data, starting point is judged as certain for the moment
It carves and proposes the corresponding data in 10 groups of data, the value for the data that every group proposes and the 20th data value before the group are done
The moment is then determined as starting point if there is the absolute value of any group of the difference is greater than 0.3 by difference, after starting point,
The 15th number after extracting the corresponding data in 10 groups of data of another moment and the values of data and the group that propose every group
Difference is done according to value, if the absolute value of all groups of difference is all larger than 0.12 and full in 80 data after another moment
The absolute value of the difference of the 15th data value then judges the moment for end point less than 0.12 after sufficient data value and the data.
5. multidimensional sign Language Recognition Method according to claim 4, which is characterized in that in the step 4, to individual character group into
Row feature extraction includes extracting specific features, algorithm according to the identification feature library of Sign Language Recognition template after being loaded into Sign Language Recognition template
Identification carries out algorithm identification using three-layer neural network according to the algorithm recognition template of Sign Language Recognition template.
6. multidimensional sign Language Recognition Method according to claim 5, which is characterized in that in the step 4, intensively refer to
Absolute value of the myoelectricity data at certain time point after normalization be more than 0.12 be known as available point, if available point is at continuous 10
In data point number be more than 5 then 10 points be referred to as valid data section, if multiple valid data sections be it is uninterrupted be connected, this is more
A valid data section connects the data segment to be formed and is known as intensive appearance.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811403415.8A CN109656358A (en) | 2018-11-23 | 2018-11-23 | A kind of multidimensional sign Language Recognition Method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811403415.8A CN109656358A (en) | 2018-11-23 | 2018-11-23 | A kind of multidimensional sign Language Recognition Method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109656358A true CN109656358A (en) | 2019-04-19 |
Family
ID=66112131
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811403415.8A Pending CN109656358A (en) | 2018-11-23 | 2018-11-23 | A kind of multidimensional sign Language Recognition Method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109656358A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110286774A (en) * | 2019-07-03 | 2019-09-27 | 中国科学技术大学 | A kind of sign Language Recognition Method based on Wrist-sport sensor |
CN110457673A (en) * | 2019-06-25 | 2019-11-15 | 北京奇艺世纪科技有限公司 | A kind of natural language is converted to the method and device of sign language |
WO2021008320A1 (en) * | 2019-07-18 | 2021-01-21 | 腾讯科技(深圳)有限公司 | Sign language recognition method and apparatus, computer-readable storage medium, and computer device |
CN114677757A (en) * | 2022-03-18 | 2022-06-28 | 吉林云帆智能工程有限公司 | Running sign language recognition algorithm for rail vehicle |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104134060A (en) * | 2014-08-03 | 2014-11-05 | 上海威璞电子科技有限公司 | Sign language interpreting, displaying and sound producing system based on electromyographic signals and motion sensors |
CN105204611A (en) * | 2014-06-24 | 2015-12-30 | 北京创思博德科技有限公司 | Finger movement mode based Morse code character inputting system and method |
CN105205436A (en) * | 2014-06-03 | 2015-12-30 | 北京创思博德科技有限公司 | Gesture identification system based on multiple forearm bioelectric sensors |
CN105446461A (en) * | 2014-06-24 | 2016-03-30 | 北京创思博德科技有限公司 | Gesture recognition method, palm virtual keyboard using same, and input method |
CN108196668A (en) * | 2017-12-05 | 2018-06-22 | 重庆中电大宇卫星应用技术研究所 | A kind of portable gesture recognition system and method |
CN108766434A (en) * | 2018-05-11 | 2018-11-06 | 东北大学 | A kind of Sign Language Recognition translation system and method |
-
2018
- 2018-11-23 CN CN201811403415.8A patent/CN109656358A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105205436A (en) * | 2014-06-03 | 2015-12-30 | 北京创思博德科技有限公司 | Gesture identification system based on multiple forearm bioelectric sensors |
CN105204611A (en) * | 2014-06-24 | 2015-12-30 | 北京创思博德科技有限公司 | Finger movement mode based Morse code character inputting system and method |
CN105446461A (en) * | 2014-06-24 | 2016-03-30 | 北京创思博德科技有限公司 | Gesture recognition method, palm virtual keyboard using same, and input method |
CN104134060A (en) * | 2014-08-03 | 2014-11-05 | 上海威璞电子科技有限公司 | Sign language interpreting, displaying and sound producing system based on electromyographic signals and motion sensors |
CN108196668A (en) * | 2017-12-05 | 2018-06-22 | 重庆中电大宇卫星应用技术研究所 | A kind of portable gesture recognition system and method |
CN108766434A (en) * | 2018-05-11 | 2018-11-06 | 东北大学 | A kind of Sign Language Recognition translation system and method |
Non-Patent Citations (1)
Title |
---|
曹翔 等: "基于fHMM分类优化的多传感器手语手势识别方法", 《航天医学与医学工程》 * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110457673A (en) * | 2019-06-25 | 2019-11-15 | 北京奇艺世纪科技有限公司 | A kind of natural language is converted to the method and device of sign language |
CN110457673B (en) * | 2019-06-25 | 2023-12-19 | 北京奇艺世纪科技有限公司 | Method and device for converting natural language into sign language |
CN110286774A (en) * | 2019-07-03 | 2019-09-27 | 中国科学技术大学 | A kind of sign Language Recognition Method based on Wrist-sport sensor |
CN110286774B (en) * | 2019-07-03 | 2021-08-13 | 中国科学技术大学 | Sign language identification method based on wrist motion sensor |
WO2021008320A1 (en) * | 2019-07-18 | 2021-01-21 | 腾讯科技(深圳)有限公司 | Sign language recognition method and apparatus, computer-readable storage medium, and computer device |
US11749029B2 (en) | 2019-07-18 | 2023-09-05 | Tencent Technology (Shenzhen) Company Limited | Gesture language recognition method and apparatus, computer-readable storage medium, and computer device |
CN114677757A (en) * | 2022-03-18 | 2022-06-28 | 吉林云帆智能工程有限公司 | Running sign language recognition algorithm for rail vehicle |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109656358A (en) | A kind of multidimensional sign Language Recognition Method | |
CN104850825B (en) | A kind of facial image face value calculating method based on convolutional neural networks | |
CN102682302B (en) | Human body posture identification method based on multi-characteristic fusion of key frame | |
CN111291678B (en) | Face image clustering method and device based on multi-feature fusion | |
Kumar et al. | Sign language recognition | |
CN106096662B (en) | Human motion state identification based on acceleration transducer | |
CN116226691B (en) | Intelligent finger ring data processing method for gesture sensing | |
CN109815826A (en) | The generation method and device of face character model | |
CN107678550A (en) | A kind of sign language gesture recognition system based on data glove | |
CN104517097A (en) | Kinect-based moving human body posture recognition method | |
JP2017516176A (en) | Method and system for recognizing handprints and gestures | |
CN108305283A (en) | Human bodys' response method and device based on depth camera and basic form | |
CN105930770B (en) | A kind of human motion recognition method based on Gaussian process latent variable model | |
CN105138974B (en) | A kind of multi-modal Feature fusion of finger based on Gabor coding | |
CN115294658B (en) | Personalized gesture recognition system and gesture recognition method for multiple application scenes | |
CN103824051A (en) | Local region matching-based face search method | |
CN109101864A (en) | The upper half of human body action identification method returned based on key frame and random forest | |
CN108764282A (en) | A kind of Class increment Activity recognition method and system | |
EP2892007A2 (en) | Static posture based person identification | |
CN110796101A (en) | Face recognition method and system of embedded platform | |
CN106264545B (en) | Step recognition method and device | |
CN106599785A (en) | Method and device for building human body 3D feature identity information database | |
CN101561881B (en) | Emotion identification method for human non-programmed motion | |
CN111126280A (en) | Gesture recognition fusion-based aphasia patient auxiliary rehabilitation training system and method | |
CN110929558A (en) | Pedestrian re-identification method based on deep learning |
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: 20190419 |