CN104036251A - Method for recognizing gestures on basis of embedded Linux system - Google Patents
Method for recognizing gestures on basis of embedded Linux system Download PDFInfo
- Publication number
- CN104036251A CN104036251A CN201410276960.0A CN201410276960A CN104036251A CN 104036251 A CN104036251 A CN 104036251A CN 201410276960 A CN201410276960 A CN 201410276960A CN 104036251 A CN104036251 A CN 104036251A
- Authority
- CN
- China
- Prior art keywords
- gesture
- linux system
- picture
- training
- classifier
- 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
Landscapes
- Image Analysis (AREA)
Abstract
The invention provides a method for recognizing gestures on the basis of an embedded Linux system. The method includes steps of firstly, porting a u-boot (universal boot loader) in an embedded core board, tailoring and porting a Linux system, making a root file system and building the embedded Linux system; secondly, acquiring pictures of at least one gesture by the aid of an image acquisition portion; thirdly, converting the pictures into gray images by the aid of an image processing portion, suppressing noise of the pictures and sharpening the pictures to obtain preprocessed pictures; fourthly, extracting features of the preprocessed pictures to obtain seven matrixes, using four front matrixes among the seven matrixes as feature parameters to form a four-dimensional feature vector and using the four-dimensional feature vector as a training datum for a rear classifier; fifthly, using the extracted feature vector as a sample datum, inputting the sample datum into a support vector machine, training, testing and scoring the support vector machine and selectively utilizing classifications corresponding to sub-classifiers with the highest scores as gesture classifications. The method has the advantage that the gesture recognition rate can be increased by the aid of the method.
Description
Technical field
The present invention relates to gesture identification field, especially relate in this field, can improve gesture identification discrimination based on embedded Linux system gesture identification method.
Background technology
Gesture identification, as a kind of interactive mode intuitively naturally, has become a large study hotspot of field of human-computer interaction in the last few years.Wherein, the gesture identification based on computer vision is because its cost compare is cheap, operates more conveniently, more and more by people, extensively used, and this is the developing direction of following gesture identification.Yet the degree of freedom of staff and elasticity, background etc. have been brought great difficulty to gesture identification, how to overcome these difficulties to realize better the key subjects that gesture identification is current people's research.
Built-in Linux is to take Linux to be the embedded operating system on basis, and it is widely used in the fields such as mobile phone, personal digital assistant (PDA), media player, consumption electronic products and Aero-Space.
Two critical problems of gesture identification are the selection of gesture feature and the design of sorter.More conventional feature has Hu square, Zernike square, H Fourier profile square etc., and wherein, Zernike is not too responsive for noise, is generally applied to the recovery aspect of image; Although Fourier descriptor has good profile descriptive power, too responsive for details, easily cause mistake to be known; And Hu square has translation, Invariant to rotation and scale.Gesture Recognition Algorithm study general based on computer vision has template matching method, neural network, support vector machine method etc.Support vector machine is a kind of new mode method based on structural risk minimization, in solving small capital, non-linear and higher-dimension pattern recognition problem, has many distinctive advantages.
Support vector machine (svm classifier device) is as a kind of trainable machine learning method.Support vector machine (svm classifier device) method is a kind of new method proposing in recent years.The main thought of support vector machine (svm classifier device) may be summarized to be 2 points: (1) it is to analyze for linear separability situation, situation for linearly inseparable, by using non-linear map that the sample of low-dimensional input space linearly inseparable is converted into high-dimensional feature space, make its linear separability, thereby make high-dimensional feature space adopt linear algorithm to carry out linear analysis to the nonlinear characteristic of sample, become possibility.
But do not exist at present a kind of discrimination that can improve gesture identification that built-in Linux, Hu square and support vector machine are combined based on embedded Linux system gesture identification method.
Summary of the invention
For the problems referred to above, the object of the present invention is to provide a kind of discrimination that can improve gesture identification that built-in Linux, Hu square and support vector machine are combined based on embedded Linux system gesture identification method.
A kind of discrimination that can improve gesture identification provided by the invention based on embedded Linux system gesture identification method, there are following steps: step 1 for by carrying out u-boot transplanting in embedded core board, carry out linux system cutting and transplanting, carry out root file system making and build embedded Linux system; Step 2 is for collecting the picture of at least one gesture by image acquisition portion; Step 3 is for being converted to picture gray level image and picture is carried out to noise reduction and sharpening processing formation pre-service picture by image processing part; Step 4 obtains seven squares for pre-service picture is carried out to feature extraction, and using front four squares in seven squares as characteristic parameter, for image f (i, j), its (p+q) rank centre distance is
Normalized center square is η
pq=μ
pq/ μ
00 r, r=(p+q)/2+1 wherein
h
1=η
20+η
02
h
2=(η
20-η
02)
2+4η
11 2
h
3=(η
30-3η
12)
2+(3η
21-η
03)
2
H
4=(η
30+ η
12)
2+ (η
21+ η
03)
2by h
1, h
2, h
3and h
4form a four-dimensional proper vector, as the training data of sorter below, and step 5 is for to be input to the proper vector of extracting as sample data in support vector machine (svm classifier device), when the gesture of predetermined quantity N is classified, the gesture of predetermined quantity N (N>=2) is carried out to pairwise classification and need construct [N * (N-1)]/2 times to the sub-classifier of the predetermined quantity N of gesture, when structure first category and other sub-classifier of Equations of The Second Kind, choose and belong to first category and other sample data of Equations of The Second Kind as training sample, and the data sample that belongs to first category is labeled as to " 1 ", belong to other data markers of Equations of The Second Kind " 0 ", for any one gesture sampling N all over (N > 0 and be even number) formation sample data, get front N/2 composing training collection, rear N/2 forms test set, when the corresponding corresponding sub-classifier of structure gesture, select the sample data corresponding with this gesture and the training set corresponding with this gesture, and selected sample data and training set are compared, train corresponding sub-classifier, trained and used the test set corresponding with this gesture to carry out testing evaluation to corresponding sub-classifier afterwards, select classification that sub-classifier that score is the highest is corresponding as the classification of gesture.
A kind of discrimination that can improve gesture identification that invention provides based on embedded Linux system gesture identification method, also there are following steps: by touching display part, can embedded Linux system be arranged with gesture identification and be operated.
A kind of discrimination that can improve gesture identification that invention provides based on embedded Linux system gesture identification method, also there are following steps: the picture collecting is transformed into gray space to obtain corresponding gray level image, utilize medium filtering to carry out noise reduction process to gray level image, gray level image is carried out to sharpening processing, and the edge contour of outstanding gray level image and minutia form pre-service picture.
A kind of discrimination that can improve gesture identification that invention provides based on embedded Linux system gesture identification method, also there are following steps: what pre-service picture extraction feature was mainly extracted is the gesture global characteristics based on Hu square, the Hu square that each gesture N training sample is extracted to hand images forms four-dimensional proper vector, as training data.
Invention effect
According to involved in the present invention a kind of based on embedded Linux system gesture identification method, by carry out u-boot transplanting in embedded core board, carry out linux system cutting and transplanting, carry out root file system making and build embedded Linux system, image acquisition portion is connected with embedded Linux system respectively with image processing part, by image acquisition portion, collect the picture of at least one gesture, image processing part is converted to picture gray level image and picture is carried out to noise reduction and sharpening processing formation pre-service picture, then pre-service picture is carried out to feature extraction and obtain seven squares, using front four the described squares in described seven squares as characteristic parameter, form a four-dimensional proper vector, finally the proper vector of extracting is input in support vector machine (svm classifier device), support vector machine (svm classifier device) is carried out to training and testing scoring, select classification that sub-classifier that score is the highest is corresponding as the classification of gesture.Therefore when carrying out gesture identification, improved the discrimination of gesture identification.
Accompanying drawing explanation
Fig. 1 is the present invention's structured flowchart based on embedded Linux system gesture recognition system in an embodiment;
Fig. 2 is the present invention's action flow chart based on embedded Linux system gesture identification method in an embodiment;
Fig. 3 is the present invention's structural drawing based on embedded Linux system gesture recognition system in an embodiment; And
Fig. 4 is the present invention's gesture identification program flow diagram in an embodiment.
Embodiment
Referring to accompanying drawing and embodiment to involved in the present invention being explained in detail based on embedded Linux system gesture identification method.
Embodiment
Fig. 1 is the present invention's structured flowchart based on embedded Linux system gesture recognition system in an embodiment.
As shown in Figure 1, based on embedded Linux system gesture recognition system 100, comprise image acquisition portion 1, image processing part 2, feature extraction portion 3, svm classifier device 4, touch display part 5 and embedded core board 6.
Image acquisition portion 1 adopts USB camera for gathering the picture of gesture.
Image processing part 2 is converted to picture gray level image and picture is carried out to noise reduction and sharpening processing formation pre-service picture.
Feature extraction portion 3 extracts seven squares of the gesture global characteristics based on Hu square, and using front four squares as characteristic parameter, by formula, calculates four-dimensional proper vector, as the training data of sorter below.
Proper vector is input in svm classifier device 4 for svm classifier device 4 is trained, tested and classifies.
Touch display part 5 for embedded Linux system being arranged and gesture identification operation.
Embedded core board 6 adopts the Tiny6410 development board of the arm of friendliness, is connected respectively with image acquisition portion 1, image processing part 2, feature extraction portion 3, svm classifier device 4 and touch display part 5.
Fig. 2 is the present invention's action flow chart based on embedded Linux system gesture identification method in an embodiment.
Step S1-1:
By carrying out u-boot transplanting in Tiny6410 development board, carry out linux system cutting and transplanting, carry out root file system and make and build described embedded Linux system, enter step S1-2.
Step S1-2:
By the picture of USB camera collection gesture, enter step S1-3.
Step S1-3:
The picture collecting is transformed into gray space to obtain corresponding gray level image, utilize medium filtering to carry out noise reduction process to gray level image, gray level image is carried out to sharpening processing, and the edge contour of outstanding gray level image and minutia form pre-service picture, enter step S1-4.
Step S1-4:
Pre-service picture is carried out to feature extraction and obtain seven squares, using front four squares in seven squares as characteristic parameter, for image f (i, j), its (p+q) rank centre distance is
Normalized center square is η
pq=μ
pq/ μ
00 r, r=(p+q)/2+1 wherein
h
1=η
20+η
02
h
2=(η
20-η
02)
2+4η
112
h
3=(η
30-3η
12)
2+(3η
21-η
03)
2
H
4=(η
30+ η
12)
2+ (η
21+ η
03)
2by h
1, h
2, h
3and h
4form a four-dimensional proper vector, as the training data of sorter below, enter step S1-5.
Step S1-5:
Using the proper vector of extracting as sample data, be input in support vector machine (svm classifier device 4), adopt " one to one " many sorting techniques, when 5 gestures are classified, 5 gestures are carried out to pairwise classification and need construct 10 sub-classifiers, when structure first category and other sub-classifier of Equations of The Second Kind, choose and belong to first category and other sample data of Equations of The Second Kind as training sample, and the data sample that belongs to first category is labeled as to " 1 ", belong to other data markers of Equations of The Second Kind " 0 ", for any one gesture sampling, form sample data 30 times, get front 15 composing training collection, latter 15 form test set, when the corresponding corresponding sub-classifier of structure gesture, select the sample data corresponding with this gesture and the training set corresponding with this gesture, and selected sample data and training set are compared, train corresponding sub-classifier, trained and used the test set corresponding with this gesture to carry out testing evaluation to corresponding sub-classifier afterwards, select classification that sub-classifier that score is the highest is corresponding as the classification of gesture.
Fig. 3 is the present invention's structural drawing based on embedded Linux system gesture recognition system in an embodiment.
By touch screen operation interface, start video capture device collection gesture the picture collecting is processed and obtained pre-service picture, finally carry out gesture identification.
Fig. 4 is the present invention's gesture identification program flow diagram in an embodiment.
Step S2-1:
Picture pre-service, enters step S2-2.
Step S2-2:
If carry out gesture identification, enter step S2-3; If do not carry out gesture identification, enter step S2-4.
Step S2-4:
Hu Moment Feature Extraction, enters step S2-5.
S2-5:
Storage Hu moment characteristics, enters step S2-6.
Step S2-6:
Svm classifier device 4 is carried out to training and testing classification, enter step S2-7.
Step S2-7:
Determine gesture.
Step S2-3:
Hu Moment Feature Extraction, enters step S2-6.
Step S2-6:
Svm classifier device 4 is carried out to training and testing classification, enter step S2-7.
Step S2-7:
Determine gesture.
The effect of embodiment and effect
According to involved in the present invention a kind of based on embedded Linux system gesture identification method, by carry out u-boot transplanting in embedded core board, carry out linux system cutting and transplanting, carry out root file system making and build embedded Linux system, image acquisition portion is connected with embedded Linux system respectively with image processing part, by image acquisition portion, collect the picture of at least one gesture, image processing part is converted to picture gray level image and picture is carried out to noise reduction and sharpening processing formation pre-service picture, then pre-service picture is carried out to feature extraction and obtain seven squares, using front four the described squares in described seven squares as characteristic parameter, form a four-dimensional proper vector, finally the proper vector of extracting is input in support vector machine (svm classifier device), support vector machine (svm classifier device) is carried out to training and testing scoring, select classification that sub-classifier that score is the highest is corresponding as the classification of gesture.Can also embedded Linux system be arranged with gesture identification and be operated by touching display part.Therefore when carrying out gesture identification, improved the discrimination of gesture identification.
Above-described embodiment is preferred case of the present invention, is not used for limiting the scope of the invention.
Claims (4)
1. can improve gesture identification discrimination based on an embedded Linux system gesture identification method, there are following steps:
Step 1, by carrying out u-boot transplanting, carry out linux system cutting and transplanting, carry out root file system and make and build described embedded Linux system in embedded core board;
Step 2, collects the picture of gesture described at least one by image acquisition portion;
Step 3, is converted to described picture gray level image and picture is carried out to noise reduction and sharpening processing formation pre-service picture by image processing part;
Step 4, carries out feature extraction to described pre-service picture and obtains seven squares, and using front four the described squares in described seven squares as characteristic parameter, for image f (i, j), its (p+q) rank centre distance is
Normalized center square is η
pq=μ
pq/ μ
00 r, r=(p+q)/2+1 wherein
h
1=η
20+η
02
h
2=(η
20-η
02)
2+4η
11 2
h
3=(η
30-3η
12)
2+(3η
21-η
03)
2
H
4=(η
30+ η
12)
2+ (η
21+ η
03)
2by h
1, h
2, h
3and h
4form a four-dimensional proper vector, as the training data of sorter below; And
Step 5, using the described proper vector of extracting as sample data, be input in support vector machine 15 (svm classifier device), when the described gesture of predetermined quantity N is classified, the described gesture of described predetermined quantity N (N >=2) is carried out to pairwise classification and need construct [N * (N-1)]/2 times to the sub-classifier of the described predetermined quantity N of described gesture, when structure first category and other sub-classifier of Equations of The Second Kind, choose and belong to first category and other sample data of Equations of The Second Kind as training sample, and the data sample that belongs to first category is labeled as to " 1 ", belong to other data markers of Equations of The Second Kind " 0 ", for gesture sampling N described in any one all over (N > 0 and be even number) formation sample data, get front N/2 composing training collection, rear N/2 forms test set, when the corresponding described corresponding sub-classifier of the described gesture of structure, select the described sample data corresponding with this gesture and the described training set corresponding with this gesture, and selected described sample data and described training set are compared, train described corresponding sub-classifier, trained and used the described test set corresponding with this gesture to carry out testing evaluation to described corresponding sub-classifier afterwards, select classification that described sub-classifier that score is the highest is corresponding as the classification of described gesture.
2. according to claim 1ly based on embedded Linux system gesture identification method, it is characterized in that also thering are following steps:
By touching display part, can described embedded Linux system be arranged with gesture identification and be operated.
3. according to claim 1ly based on embedded Linux system gesture identification method, it is characterized in that also thering are following steps:
The described picture collecting is transformed into gray space to obtain corresponding described gray level image, utilize medium filtering to carry out noise reduction process to described gray level image, described gray level image is carried out to sharpening processing, and edge contour and the minutia of outstanding described gray level image form described pre-service picture.
4. according to claim 1ly based on embedded Linux system gesture identification method, it is characterized in that also thering are following steps:
What described pre-service picture extraction feature was mainly extracted is the gesture global characteristics based on Hu square, and the Hu square that each gesture N training sample is extracted to hand images forms four-dimensional proper vector, as described training data.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410276960.0A CN104036251A (en) | 2014-06-20 | 2014-06-20 | Method for recognizing gestures on basis of embedded Linux system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410276960.0A CN104036251A (en) | 2014-06-20 | 2014-06-20 | Method for recognizing gestures on basis of embedded Linux system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN104036251A true CN104036251A (en) | 2014-09-10 |
Family
ID=51467017
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410276960.0A Pending CN104036251A (en) | 2014-06-20 | 2014-06-20 | Method for recognizing gestures on basis of embedded Linux system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104036251A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108806375A (en) * | 2018-05-09 | 2018-11-13 | 河南工学院 | A kind of interactive teaching method and system based on image recognition |
CN111382598A (en) * | 2018-12-27 | 2020-07-07 | 北京搜狗科技发展有限公司 | Identification method and device and electronic equipment |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102339379A (en) * | 2011-04-28 | 2012-02-01 | 重庆邮电大学 | Gesture recognition method and gesture recognition control-based intelligent wheelchair man-machine system |
US20120068914A1 (en) * | 2010-09-20 | 2012-03-22 | Kopin Corporation | Miniature communications gateway for head mounted display |
-
2014
- 2014-06-20 CN CN201410276960.0A patent/CN104036251A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120068914A1 (en) * | 2010-09-20 | 2012-03-22 | Kopin Corporation | Miniature communications gateway for head mounted display |
CN102339379A (en) * | 2011-04-28 | 2012-02-01 | 重庆邮电大学 | Gesture recognition method and gesture recognition control-based intelligent wheelchair man-machine system |
Non-Patent Citations (2)
Title |
---|
徐成等: ""一种基于嵌入式系统实时交互的手势识别方法"", 《计算机应用研究》 * |
董立峰: ""基于Hu矩和支持向量机的静态手势识别及应用"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108806375A (en) * | 2018-05-09 | 2018-11-13 | 河南工学院 | A kind of interactive teaching method and system based on image recognition |
CN111382598A (en) * | 2018-12-27 | 2020-07-07 | 北京搜狗科技发展有限公司 | Identification method and device and electronic equipment |
CN111382598B (en) * | 2018-12-27 | 2024-05-24 | 北京搜狗科技发展有限公司 | Identification method and device and electronic equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103065134B (en) | A kind of fingerprint identification device and method with information | |
CN102938065B (en) | Face feature extraction method and face identification method based on large-scale image data | |
CN106203356B (en) | A kind of face identification method based on convolutional network feature extraction | |
CN109451634B (en) | Gesture-based electric lamp control method and intelligent electric lamp system thereof | |
CN105956560A (en) | Vehicle model identification method based on pooling multi-scale depth convolution characteristics | |
CN103258157B (en) | A kind of online handwriting authentication method based on finger information and system | |
US9552509B2 (en) | Method and system for rectifying distorted fingerprint | |
CN109145766A (en) | Model training method, device, recognition methods, electronic equipment and storage medium | |
CN102467657A (en) | Gesture recognizing system and method | |
Bharadi et al. | Off-line signature recognition systems | |
CN104134061A (en) | Number gesture recognition method for support vector machine based on feature fusion | |
CN105740808B (en) | Face identification method and device | |
CN105117708A (en) | Facial expression recognition method and apparatus | |
CN104834905A (en) | Facial image identification simulation system and method | |
CN102542243A (en) | LBP (Local Binary Pattern) image and block encoding-based iris feature extracting method | |
CN102622590A (en) | Identity recognition method based on face-fingerprint cooperation | |
CN101930549A (en) | Second generation curvelet transform-based static human detection method | |
CN102663401A (en) | Image characteristic extracting and describing method | |
Kumar et al. | A hybrid gesture recognition method for American sign language | |
CN104036245B (en) | A kind of biological feather recognition method based on online Feature Points Matching | |
CN103984954B (en) | Image combining method based on multi-feature fusion | |
CN103839066A (en) | Feature extraction method based on biological vision | |
CN101520839B (en) | Human body detection method based on second-generation strip wave conversion | |
CN110866468A (en) | Gesture recognition system and method based on passive RFID | |
CN106650696A (en) | Handwritten electrical element identification method based on singular value decomposition |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into 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: 20140910 |