CN109034093A - A kind of design and realization of quick dynamic Sign Language Recognition algorithm - Google Patents
A kind of design and realization of quick dynamic Sign Language Recognition algorithm Download PDFInfo
- Publication number
- CN109034093A CN109034093A CN201810906565.4A CN201810906565A CN109034093A CN 109034093 A CN109034093 A CN 109034093A CN 201810906565 A CN201810906565 A CN 201810906565A CN 109034093 A CN109034093 A CN 109034093A
- Authority
- CN
- China
- Prior art keywords
- gesture
- module
- sign language
- realization
- design
- 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
- 238000013461 design Methods 0.000 title claims abstract description 14
- 238000012545 processing Methods 0.000 claims abstract description 9
- 238000001914 filtration Methods 0.000 claims abstract description 4
- 238000004891 communication Methods 0.000 claims description 4
- 238000005452 bending Methods 0.000 claims description 2
- 230000015572 biosynthetic process Effects 0.000 claims description 2
- 238000003786 synthesis reaction Methods 0.000 claims description 2
- 238000013519 translation Methods 0.000 claims description 2
- 230000005540 biological transmission Effects 0.000 claims 1
- 230000010354 integration Effects 0.000 claims 1
- 238000000034 method Methods 0.000 abstract description 17
- 230000006870 function Effects 0.000 abstract description 8
- 238000012549 training Methods 0.000 abstract description 7
- 230000004927 fusion Effects 0.000 abstract description 3
- 238000012360 testing method Methods 0.000 abstract description 3
- 230000006978 adaptation Effects 0.000 abstract description 2
- 238000005457 optimization Methods 0.000 description 4
- 230000001186 cumulative effect Effects 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 206010011878 Deafness Diseases 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000003183 myoelectrical effect Effects 0.000 description 2
- PCTMTFRHKVHKIS-BMFZQQSSSA-N (1s,3r,4e,6e,8e,10e,12e,14e,16e,18s,19r,20r,21s,25r,27r,30r,31r,33s,35r,37s,38r)-3-[(2r,3s,4s,5s,6r)-4-amino-3,5-dihydroxy-6-methyloxan-2-yl]oxy-19,25,27,30,31,33,35,37-octahydroxy-18,20,21-trimethyl-23-oxo-22,39-dioxabicyclo[33.3.1]nonatriaconta-4,6,8,10 Chemical compound C1C=C2C[C@@H](OS(O)(=O)=O)CC[C@]2(C)[C@@H]2[C@@H]1[C@@H]1CC[C@H]([C@H](C)CCCC(C)C)[C@@]1(C)CC2.O[C@H]1[C@@H](N)[C@H](O)[C@@H](C)O[C@H]1O[C@H]1/C=C/C=C/C=C/C=C/C=C/C=C/C=C/[C@H](C)[C@@H](O)[C@@H](C)[C@H](C)OC(=O)C[C@H](O)C[C@H](O)CC[C@@H](O)[C@H](O)C[C@H](O)C[C@](O)(C[C@H](O)[C@H]2C(O)=O)O[C@H]2C1 PCTMTFRHKVHKIS-BMFZQQSSSA-N 0.000 description 1
- 230000003542 behavioural effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005315 distribution function Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000001771 impaired effect Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000009711 regulatory function Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 230000001755 vocal effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
- G06V40/28—Recognition of hand or arm movements, e.g. recognition of deaf sign language
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
- G06V20/47—Detecting features for summarising video content
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Human Computer Interaction (AREA)
- Image Analysis (AREA)
Abstract
The method of the present invention belongs to Intellisense field.Specially a kind of design implementation method of quick dynamic Sign Language Recognition algorithm.The hardware platform that the present invention realizes is using STM32F103RCT6 minimum system as processor, and as platform, ARM built in hardware platform is handled μ COS system.The invention firstly uses Baum-Welch algorithms repeatedly to be trained, for the independent HMM model parameter of each gesture training, obtain stable test template, the end PC is sent to by WIFI, then sign language data template being copied to can be inserted on MCU in SDIO mouthfuls of SD card, then start to acquire all data by each sensor, it is transferred to micro treatment module, by data fusion, filtering processing, it is transferred in primary processor by WiFi module, carry out data processing, it is packaged, the gesture template in SD card is read again, finally adaptation function is called to carry out gesture identification using DTW and HMM algorithm, end of identification returns to the functional value for representing gesture sign language, on the one hand display text information, text information is subjected to voice broadcast in voice synthetic module on the other hand, it is whole to realize A identification process.
Description
Technical field
The method of the present invention belongs to Intellisense field.Particular by optimization, reconstruct DTW and HMM algorithm, realization is more adapted to
The algorithm of lower time complexity and more change in time and space ability rapidly and accurately completes hand to realize in the dynamic case
Language identification.
Background technique
With computing capability raising and science and technology development, Gesture Recognition using increasingly extensive.Traditional
When in face of the situation that classification information amount is big, environment is complex, recognition effect is often difficult satisfactory Gesture Recognition.
It is shown according to the data that the generaI investigation of newest China human mortality is announced, is reached more than 8500 in our all kinds of disabled person's sums of country
Ten thousand people, wherein hearing disabilities number is about 20,540,000, and speech disabilities number is about 1,300,000, accounts for national disabled person's sum respectively
24.16% and 1.53%.In terms of exchange, there are the deaf-mutes of IMPAIRED VERBAL COMMUNICATION to be handed over accordingly by sign language
Stream, there are great obstacles for the communication of they and normal person.But according to existing market investigate find, on the market presently, there are
Sign language interpreter equipment, it is most of be by using camera or surface myoelectric sensor to hand Posture acquisition user
Behavioral data.However, there is processing using the equipment of camera data volume be huge, visual dead angle and it is inconvenient to carry the disadvantages of;It adopts
Then expensive with surface myoelectric sensor device, most deaf-mutes can not undertake.In fact, this is exactly one kind by palm in fact
The dynamic data that upper each site sensor obtains, by a kind of sophisticated identification algorithm combined carry out parsing sign language instruction again with
Gesture library is matched, then by signal by converting and then obtaining voice or text information, difficult point is just in dynamic Sign Language Recognition
On algorithm.
The feature of dynamic gesture is mainly track, and track is the sequence of spatial position coordinate pair time, and the same gesture exists
Therefore possible cause different in size in time series it is different to can be very good solution time span using dynamic time warping algorithm
The problem of cause.
Dynamic time warping (DTW) is a very famous algorithm, is applied to field of speech recognition in early days, up to now
Through being widely used in multiple fields.The concept of time series has application in many fields, and DTW is thought based on Dynamic Programming
Think, can be used for the similarity calculation between the inconsistent time series of length.Distortion measurement in DTW algorithm can apply Europe
Family name's distance can also use log-likelihood ratio distance, and algorithm is simple, recognition correct rate is high.
Hidden Markov model (HMM) algorithm is a kind of classic algorithm of dynamic hand gesture recognition, applies know in voice earliest
Other field, after in Computer Image Processing and mobile communication, the fields such as biological information are also used widely.The algorithm is based on horse
Markov's chain model, current state cannot be observed to obtain, but the observation state before passing through is calculated, each observe to
Amount is indicated by a status switch, this status switch has probability density distribution corresponding with observation vector.Ma Erke
Husband's model parameter can use discrete probability distribution function, using maximum posteriori criterion as decision-making technique.Hidden Ma Erke
Husband's algorithm is widely used in dynamic hand gesture recognition, and real-time is good, accuracy is high.
Although DTW algorithm and HMM algorithm are applied in dynamic hand gesture recognition, the accuracy of identification and recognition speed all can
When being greatly improved, but being applied in dynamic Sign Language Recognition, calculation amount is just very huge, search space and road
Diameter is complicated, and algorithm optimization if inappropriate, then will affect the performance that arithmetic analysis instruction is realized.Especially in some cases,
Dynamic time warping (DTW) this algorithm is attempted to explain the variation in y-axis by distortion x-axis, this will lead to unnatural rule
It is whole, so that the certain point on time-serial position has been mapped to a certain section of longer section on another time-serial position.
Summary of the invention
The purpose of the method for the present invention is optimization and reconstructs a kind of recognizer, and finds a kind of system side according to this algorithm
Case, solution is computationally intensive on dynamic Sign Language Recognition algorithm at present, and resolution is inaccurate, and recognition speed is low, recognizer
Power consumption is big, low efficiency, the problem of to mobile platform bad adaptability etc..
The present invention provides based on DTW and HMM algorithm dynamic Sign Language Recognition algorithm realize design method, realization it is hard
Part platform is using STM32F103RCT6 minimum system as processor, and μ COS system is as platform, built in hardware platform at ARM
Reason.The invention firstly uses Baum-Welch algorithms repeatedly to be trained, and trains independent HMM model parameter for each gesture,
We acquire 20 frame data for each isolated sign words altogether, and frame period 30ms obtains stable test template, sent by WIFI
To the end PC, being then copied to sign language data template can be inserted on MCU in SDIO mouthfuls of SD card, then start to be adopted by each sensor
Collect all data, receives and stores data into micro treatment module, analog quantity therein is converted to digital quantity by ADC, by number
It according to fusion, filtering processing, is transferred in primary processor by WiFi module, is driven by code and carry out data processing, be packaged, then read
The gesture template in SD card is taken, adaptation function is finally called to carry out gesture identification using DTW and HMM algorithm, end of identification returns
On the one hand the functional value for representing gesture sign language shows text information on OLED, while utilizing the communication of UART on the other hand
Mode transmits information to voice synthetic module, and then carries out voice broadcast, to realize entire identification process.
In conjunction with attached drawing 1, algorithm flow of the invention is as follows:
The present invention combines the hidden horse of HMM of the DTW template matching algorithm and more change in time and space ability of lower time complexity
Er Kefu model algorithm advantage, proposes a kind of recognizer of synthesis.Template gesture library is divided using Time Series Clustering algorithm first
If reusing DTW algorithm at Ganlei and calculating input gesture and template library distance value, and be classified to the smallest classification of distance value
In, some gesture is accurately finally searched in concrete kind using HMM, thus evaded the computationally intensive problem of HMM well,
And the accuracy and recognition speed identified is all greatly improved.Specifically include following procedure:
1. by template gesture library divide into several classes
(1) by Baum-Welch algorithm, for each template gesture training HMM model parameter lambda (A, B, π) A=
{aij}N+N(1≤i, j≤N) indicates t moment state qiTurn to the q at t+1 momentjTransition probability matrix, wherein aij=P (it+1=
qj|it=qi);B is observation probability matrix: B={ bj(k)}N*M, wherein bj(k)=P (ot=vk|it=qi) it is in t moment state
qiUnder the conditions of observe vkProbability;π=(π1,π2...πN) indicate initial state probabilities set, and πi=P (ii=qi)(1≤i
≤ N) indicate that the initial t=1 moment is in state qiProbability.
In actual gesture identification translation process, the status switch V { v of gesture1,v2...vTIt is unknown, but its hand
Observation sequence O { the o of gesture1,o2...osIt is visible, observable, it is possible to use the hidden horse of Baum-Welch algorithm training
Er Kefu model parameter λ (A, B, π), and status switch data are regarded as unobservable hidden data I and are handled.Then hidden Ma Erke
Husband's model is actually the probabilistic model containing hidden variable:
P (O | λ)=∑tP(O|I,λ)P(I|λ)(I)
Therefore the realization of EM (expectation-maximization algorithm) algorithm can be used in Hidden Markov Model parameter learning.Specifically, first
Q function is found out, then maximizes to it and obtains A, B, π.
1. finding out Q function
Wherein λ is the Hidden Markov Model parameter for wanting very big value,It is currently estimating for Hidden Markov Model parameter
Evaluation.
2. the model parameter A, B, π of the Q function that maximizes
πi=γI(i), lead to
It crosses and initializes following sequencebj(k)0,Recursion obtains model parameter:
λn+1=(A(n+1),B(n+1),λ(n+1))
(2) use Time Series Clustering algorithm by template gesture library divide into several classes
2. will input gesture be included into in the smallest classification of template gesture distance value
(1) input gesture and each gesture template distance value of template library are calculated using DTW algorithm
DTW finds the regulatory function of Best Times between list entries and template sequence by dynamic programming algorithm (DP),
Keep the distance of the two most short.If template gesture time sequence Q={ q1,q2...qi...qmAnd input gesture time sequence C=
{c1,c2…cj…cn, length is respectively m and n, and definition Time alignment function is γ={ γ (1), γ (2) ... γ (X) }, and X is
Path length, warping function point is to γ (x)=(γq(x),γc(x)),γq(x)∈{1,2,…,m},γc(x)∈{1,
2 ..., n }, γ (x) indicates template gesture Q γq(x) γ of a vector and input gesture Cc(x) point pair that a vector is constituted.
Input the Euclidean distance d (γ of similarity between the two between gesture and template gestureq(x),γc(x)) it comments
Valence.Dynamic time warping algorithm idea is exactly to realize that the total distance value of two time serieses is minimum.That is:
(2) input gesture is included into the smallest classification of distance value;
As shown in Fig. 2, the matrix grid of a m × n is constructed, horizontal axis indicates that list entries, the longitudinal axis indicate template sequence.
Wherein each grid represents Euler's distance value of list entries and template sequence, uses dx(qi,cj) indicate.And have:
d(γq(x),γc(x))=dx(qi,cj)
dx(qi,cj)=(qi-cj)2
As shown in Fig. 2, it is minimum can to find out a cumulative distance value D (Q, C) from point (1,1) to point (m, n) for we
Optimal path W, wherein W=w1,w2…wk…wK, max (m, n)≤K < m+n-1 must satisfy three in path selection about
Beam condition, if current point wx=(iq,jc), previous point wx-1=(i'q,j'c), then have:
1. boundary condition: w1=(1,1) and wK=(m, n);
2. continuity: (iq-i'q)≤and (jc-j'c)≤1;
3. monotonicity: (iq-i'q) >=0 and (jc-j'c) >=0,
That is: (i'q,j'c)∈{(iq-1,jc),(iq,jc-1),(iq-1,jc}, -1) recurrence Relation is thus obtained are as follows:
Work as iq→m,jcWhen → n, D (m, n) is the total Euler's distance value inputted between gesture and template gesture.D (m, n) is more
Small, the two is more similar, conversely, more dissimilar.
Further, the path of DTW is limited, algorithm execution efficiency can be improved.In input gesture and template gesture
Timing, input gesture and the few part similarity of gesture in gesture library are high, and cumulative distance is small, i.e., in " O " type region, such as attached drawing 3
It is shown;And major part similarity is low, cumulative distance is big, in " X " type region, so identification path is limited in parallelogram
In ABCD frame, algorithm operation quantity can be greatly reduced, improve the matched efficiency of gesture.
3. searching gesture in (template) concrete kind
By Baum-Welch algorithm be each gesture training pattern parameter lambda after, when carrying out gesture matching, calculating pair
P (O | the λ) probability answered, taking the maximum value of probability is final matched gesture model.But the time that P (O | λ) is directly calculated is complicated
Degree is high, so we simplify calculating by forward algorithm (forward).Specifically, the observation sequence of note t moment is O { o1,
o2...otAnd state be qiProbability be preceding to probability αt(i), and there is αt(i)=P (o1,o2...ot,it=qi| λ), initial value
αt(i)=πibi(oI), i=1,2,3...N, by the way that t=1,2 ... T-1 recursion have:
As t=T-1, there is probability:
In gesture matching process, the probability P (O of each of vocabulary HMM model parameter lambda after DTW classifies is calculated
|λV), choosing wherein maximum probability value is optimal HMM model parameter, it is believed that its corresponding standard gesture is matched with input gesture,
And have:
Detailed description of the invention
Fig. 1 algorithmic system block diagram
Fig. 2 DTW template matching
Fig. 3 DTW algorithm optimization
Fig. 4 algorithm realizes system block diagram
Fig. 5 data samples
Fig. 6 gloves simulate structure chart
The practical gesture identification procedure chart of Fig. 7
Fig. 8 subject's Sign Language Recognition rate and recognition time result statistics
Specific embodiment
A kind of dynamic Sign Language Recognition algorithm design implementation method based on DTW and HMM algorithm.Consider that dynamic sign language instruction is known
Computationally intensive on other algorithm, resolution is inaccurate, and recognition speed is low, and recognizer power consumption is big, low efficiency etc. is asked
Topic is in conjunction with shown in attached drawing 4, and the specific implementation steps are as follows:
1. preparation: choosing experimenter and obtain gesture training sample set, then carried out repeatedly by Baum-Welch algorithm
Training obtains stable test template for the independent HMM model parameter of each gesture training.With a wherein experimental data point
Analysis, we acquire 20 frame data, frame period 30ms for each isolated sign words altogether, and are sent to computer by WIFI.Its data
Sampled value is as shown in Fig. 5.
2. receiving and storing the various kinds of sensors acquisition digital flexion degree at each position on Intelligent glove, whether finger tip connects
The data (including touch sensor, gyroscope, bending sensor) such as touching, the posture of gesture in space.Data acquisition sensing
Device is distributed shown in attached drawing 6.
3. collected gesture data by data fusion, filtering processing, is transferred to master by WiFi by micro treatment module
Processor.
4. primary processor is the μ COS system based on STM32F103RCT6, good and voice module baud rate, data are configured
The UART such as position, stop position parameter and with after the WIFI parameter such as the IP of sign language gloves, port, starts to receive and store sign language gloves
Gesture data.
5. handle data in RAM, by the template gesture storehouse matching in dynamic time warping (DTW) algorithm, with SD card,
If Data Matching success, is translated as text by return function value.On the one hand identification text is shown by OLED;Another aspect quilt
Primary processor is sent into voice module and carries out voice broadcast, to realize entire identification process.The practical gesture simulated in PC is known
Shown in other process attached drawing 7.
A kind of dynamic Sign Language Recognition algorithm design implementation method based on DTW and HMM algorithm.Subject's Sign Language Recognition rate and
Recognition time result statistics is shown in attached drawing 8.
Claims (8)
1. a kind of design and realization of quick dynamic Sign Language Recognition algorithm, it is characterised in that: gesture data acquisition module, gesture number
According to receiving module, WiFi module, template gesture library module, gesture data identification module, voice synthetic module, display module.
2. the design and realization of a kind of quick dynamic Sign Language Recognition algorithm according to claim 1, it is characterised in that described
Gesture data acquisition module is realized the acquisition of gesture data by touch sensor, gyroscope, bending sensor.
3. the design and realization of a kind of quick dynamic Sign Language Recognition algorithm according to claim 1, it is characterised in that described
Gesture data receiving module is realized reception, filtering, the storage of data by microprocessor, and wherein analog quantity will first pass through ADC module
Be converted to digital quantity.
4. the design and realization of a kind of quick dynamic Sign Language Recognition algorithm according to claim 1, it is characterised in that gesture
The transmission of the data of receiving module and gesture recognition module realizes that gesture is sent to after receiving using the WiFi module of ESP8266
Main control chip end.
5. the design and realization of a kind of quick dynamic Sign Language Recognition algorithm according to claim 4, it is characterised in that described
Main control chip use STM32F103RCT6.
6. the design and realization of a kind of quick dynamic Sign Language Recognition algorithm according to claim 1, it is characterised in that described
Gesture data identification module is that the gesture data received is passed through DTW algorithm and the progress of template gesture library by main control chip
Match, carry out accurately searching template gesture in gesture library by HMM algorithm and be translated as corresponding text information.
7. the design and realization of a kind of quick dynamic Sign Language Recognition algorithm according to claim 6, it is characterised in that described
Gesture data processing, integration, identification, translation related algorithm write on ARM Cortex-M3 central processing unit.
8. the design and realization of a kind of quick dynamic Sign Language Recognition algorithm according to claim 4, it is characterised in that
XFC5152 voice synthetic module is communicated using UART communication modes with primary processor, and then text information is carried out speech synthesis
And real-time voice plays.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810906565.4A CN109034093A (en) | 2018-08-10 | 2018-08-10 | A kind of design and realization of quick dynamic Sign Language Recognition algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810906565.4A CN109034093A (en) | 2018-08-10 | 2018-08-10 | A kind of design and realization of quick dynamic Sign Language Recognition algorithm |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109034093A true CN109034093A (en) | 2018-12-18 |
Family
ID=64633421
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810906565.4A Pending CN109034093A (en) | 2018-08-10 | 2018-08-10 | A kind of design and realization of quick dynamic Sign Language Recognition algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109034093A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109902554A (en) * | 2019-01-09 | 2019-06-18 | 天津大学 | A kind of recognition methods of the sign language based on commercial Wi-Fi |
CN112686132A (en) * | 2020-12-28 | 2021-04-20 | 南京工程学院 | Gesture recognition method and device |
CN113947683A (en) * | 2021-10-15 | 2022-01-18 | 兰州交通大学 | Fingertip point detection method and system and fingertip point motion track identification method and system |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101605399A (en) * | 2008-06-13 | 2009-12-16 | 英华达(上海)电子有限公司 | A kind of portable terminal and method that realizes Sign Language Recognition |
CN102063825A (en) * | 2010-12-15 | 2011-05-18 | 北京理工大学 | Sign language recognizing device based on data gloves |
CN103902984A (en) * | 2014-04-15 | 2014-07-02 | 福州大学 | Improved HMM training algorithm for dynamic hand gesture recognition |
CN205158728U (en) * | 2015-11-19 | 2016-04-13 | 陆庆健 | Sign language translating system |
CN106446861A (en) * | 2016-09-28 | 2017-02-22 | 辽宁石油化工大学 | Sign language recognition system, device and method |
CN107230414A (en) * | 2017-06-16 | 2017-10-03 | 天津职业技术师范大学 | A kind of SCM Based Sign Language Recognition translation system |
CN107678550A (en) * | 2017-10-17 | 2018-02-09 | 哈尔滨理工大学 | A kind of sign language gesture recognition system based on data glove |
-
2018
- 2018-08-10 CN CN201810906565.4A patent/CN109034093A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101605399A (en) * | 2008-06-13 | 2009-12-16 | 英华达(上海)电子有限公司 | A kind of portable terminal and method that realizes Sign Language Recognition |
CN102063825A (en) * | 2010-12-15 | 2011-05-18 | 北京理工大学 | Sign language recognizing device based on data gloves |
CN103902984A (en) * | 2014-04-15 | 2014-07-02 | 福州大学 | Improved HMM training algorithm for dynamic hand gesture recognition |
CN205158728U (en) * | 2015-11-19 | 2016-04-13 | 陆庆健 | Sign language translating system |
CN106446861A (en) * | 2016-09-28 | 2017-02-22 | 辽宁石油化工大学 | Sign language recognition system, device and method |
CN107230414A (en) * | 2017-06-16 | 2017-10-03 | 天津职业技术师范大学 | A kind of SCM Based Sign Language Recognition translation system |
CN107678550A (en) * | 2017-10-17 | 2018-02-09 | 哈尔滨理工大学 | A kind of sign language gesture recognition system based on data glove |
Non-Patent Citations (2)
Title |
---|
姚桂林等: "一种基于DTW/ISODATA算法的多层分类器手语识别方法", 《计算机工程与应用》 * |
张文等: "低功耗便携智能翻译手套系统", 《电子制作》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109902554A (en) * | 2019-01-09 | 2019-06-18 | 天津大学 | A kind of recognition methods of the sign language based on commercial Wi-Fi |
CN109902554B (en) * | 2019-01-09 | 2023-03-10 | 天津大学 | Sign language identification method based on commercial Wi-Fi |
CN112686132A (en) * | 2020-12-28 | 2021-04-20 | 南京工程学院 | Gesture recognition method and device |
CN113947683A (en) * | 2021-10-15 | 2022-01-18 | 兰州交通大学 | Fingertip point detection method and system and fingertip point motion track identification method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10109219B2 (en) | System and method for automated sign language recognition | |
Memmesheimer et al. | Gimme signals: Discriminative signal encoding for multimodal activity recognition | |
Alrubayi et al. | A pattern recognition model for static gestures in malaysian sign language based on machine learning techniques | |
Ren et al. | Human action recognition in smart classroom | |
CN115019240B (en) | Grading method, device and equipment for chemical experiment operation and readable storage medium | |
CN109034093A (en) | A kind of design and realization of quick dynamic Sign Language Recognition algorithm | |
Wang et al. | Sparse observation (so) alignment for sign language recognition | |
Ghotkar et al. | Dynamic hand gesture recognition using hidden Markov model by Microsoft Kinect sensor | |
Bhalla et al. | Imu2doppler: Cross-modal domain adaptation for doppler-based activity recognition using imu data | |
Bastas et al. | Air-writing recognition using deep convolutional and recurrent neural network architectures | |
Hisham et al. | Supervised learning classifiers for Arabic gestures recognition using Kinect V2 | |
Debard et al. | Learning to recognize touch gestures: Recurrent vs. convolutional features and dynamic sampling | |
CN112749646A (en) | Interactive point-reading system based on gesture recognition | |
CN110031827A (en) | A kind of gesture identification method based on ultrasonic distance measurement principle | |
Fang et al. | A SRN/HMM system for signer-independent continuous sign language recognition | |
Xu et al. | A long term memory recognition framework on multi-complexity motion gestures | |
CN110516638B (en) | Sign language recognition method based on track and random forest | |
Cohen et al. | Recognition of continuous sign language alphabet using leap motion controller | |
CN113658690A (en) | Intelligent medical guide method and device, storage medium and electronic equipment | |
KR20190121593A (en) | Sign language recognition system | |
Nahar et al. | A robust model for translating arabic sign language into spoken arabic using deep learning | |
Oliva et al. | Filipino sign language recognition for beginners using kinect | |
Yao et al. | Decision-tree-based algorithm for 3D sign classification | |
Deng et al. | A Two-step Approach based on PaHMM for the Recognition of ASL | |
Kagalkar et al. | Mobile Application Based Translation of Sign Language to Text Description in Kannada Language. |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20181218 |