CN109766838A - A kind of gait cycle detecting method based on convolutional neural networks - Google Patents
A kind of gait cycle detecting method based on convolutional neural networks Download PDFInfo
- Publication number
- CN109766838A CN109766838A CN201910026947.2A CN201910026947A CN109766838A CN 109766838 A CN109766838 A CN 109766838A CN 201910026947 A CN201910026947 A CN 201910026947A CN 109766838 A CN109766838 A CN 109766838A
- Authority
- CN
- China
- Prior art keywords
- gait
- frame
- neural networks
- convolutional neural
- gait cycle
- 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.)
- Granted
Links
- 230000005021 gait Effects 0.000 title claims abstract description 82
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 32
- 230000000007 visual effect Effects 0.000 claims abstract description 13
- 230000008569 process Effects 0.000 claims abstract description 8
- 239000000284 extract Substances 0.000 claims abstract description 5
- 238000001914 filtration Methods 0.000 claims abstract description 5
- 238000012549 training Methods 0.000 claims description 12
- 238000010606 normalization Methods 0.000 claims description 7
- 230000003321 amplification Effects 0.000 claims description 5
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 238000009432 framing Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 claims description 2
- 238000007781 pre-processing Methods 0.000 claims description 2
- 238000013139 quantization Methods 0.000 claims description 2
- 230000008859 change Effects 0.000 abstract description 4
- 238000003745 diagnosis Methods 0.000 abstract description 2
- 230000003993 interaction Effects 0.000 abstract description 2
- 238000012544 monitoring process Methods 0.000 abstract description 2
- 238000001514 detection method Methods 0.000 description 11
- 238000004458 analytical method Methods 0.000 description 4
- 238000000605 extraction Methods 0.000 description 4
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 210000002414 leg Anatomy 0.000 description 2
- 210000001699 lower leg Anatomy 0.000 description 2
- 230000003252 repetitive effect Effects 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 208000004350 Strabismus Diseases 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000037237 body shape Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 210000003141 lower extremity Anatomy 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
Abstract
The present invention provides a kind of gait cycle detecting method based on convolutional neural networks, pre-processes to gait video, including video decoding, and pedestrian contour extracts and the normalized image pretreatment operation of mass center；Train the convolutional neural networks for extracting gait cycle feature；Gait sequence of frames of video to be measured is sent into convolutional neural networks, output waveform obtains a gait cycle after filtering, through the position of the adjacent peaks and troughs of determination.This method has very strong robustness to the variation of angle change, dress ornament and belongings, it solves the problems, such as to be difficult to detect gait cycle under front and back visual angle, the method of the present invention is significant to Gait Recognition precision in complex environment is improved, it can be used for the front end in Gait Recognition, in the identification security monitoring, human-computer interaction, medical diagnosis and access control system etc..
Description
Technical field
The invention belongs to computer vision fields, and in particular to a kind of gait cycle detection side based on convolutional neural networks
Method.
Background technique
Gait Recognition can complete the receipts of data compared with living things feature recognition mode in the unwitting situation of tester
Collect the identification of work and remote identity.And gait cycle detection is an inevitable process in Gait Recognition, a tool
There is the Algorithm for gait recognition of good discrimination to be built upon on the intact gait cycle of segmentation.Again because Gait Recognition has
Have a concealed feature, the randomness of data acquisition is larger, pedestrian relative to camera direction and pedestrian dress ornament state all
It can be arbitrary, increase the difficulty of cycle detection.
The development of gait cycle detection technique along with Gait Recognition development.In existing method mostly using pedestrian's width as
Feature carries out gait cycle detection.Document (Silhouette-Based Human Identification from Body
Shape and Gait.IEEE International Conference on Automatic Face and Gesture
Recognition.2002,366-372 it) proposes earlier and carries out period inspection using the width and altitude feature of physical trait
It surveys, but this method is affected by pedestrian and video camera distance.Document (Gait recognition with transient
binary patterns.Visual Communication and Image Representation.2015,33(C),69-
Etc. 77) propose on this basis using normalized single profile width feature progress gait cycle detection, but this method exists
Work is difficult under front and back side visual angle.Document (Silhouette Analysis-Based Gait Recognition for
Human Identification.Pattern Analysis&Machine Intelligence IEEE Transactions
On, 2003,25 (12): 1505-1518) it proposes to carry out gait cycle detection using the ratio of width to height of gait profile, avoid height
Normalization is influenced caused by pedestrian's width.Document (Human Identification Using Temporal
Information Preserving Gait Template.IEEE Transactions on Pattern Analysis&
Machine Intelligence, 2012,34 (11): 2164-76) by the extraction to each frame image lower limb mean breadth come
Frame position representative in a complete gait sequence is represented, influence of the object to pedestrian's width is avoided carrying around.It is overall
For gait cycle can effectively be detected at 90 ° of side based on body width characteristics method, but regarded in a front surface and a side surface
Error is very big when angle, or even cannot work.Document (The humanID gait challenge problem:data sets,
performance,and analysis.IEEE Transactions on Pattern Analysis&Machine
Intelligence, 2005,27 (2): 162-177) Selection utilization human body two-dimensional areas as cycle detection feature carry out
Feature extraction.But error very big problem when equally existing a front surface and a side surface visual angle.Document (Dual-ellipse fitting
approach for robust gait periodicity detection.Neurocomputing,2012,79(3):173-
178) a kind of method based on bielliptic(al) fitting is proposed.With center of mass point be divide pedestrian central point by pedestrian contour according to figure
The vertical direction of picture is divided into left and right two halves, is fitted respectively with ellipse, so that pedestrian contour is tangent with ellipse, it is ellipse to calculate this
Round eccentricity, the elliptical eccentricity of two fittings are added the periodic feature that the frame image is indicated as feature.In side
Higher discrimination is achieved under 90 ° of face, front and back visual angle, but larger in such as 18 ° of angle of strabismus and 36 ° of equal errors.
Deep learning is fast-developing in recent years, and convolutional neural networks are wide as a kind of effective image characteristics extraction tool
It is general to apply to computer vision field.It is inspired by it, set forth herein a kind of gait cycle detection side based on convolutional neural networks
Method extracts the gait cycle feature of each frame by training convolutional neural networks, using this feature location present frame in step
Gait cycle detection process is completed in position in the state period.This method has stronger robustness, in different perspectives, different
More accurate gait cycle can be detected in the case where dress and belongings.
Summary of the invention
The object of the present invention is to provide a kind of gait cycle detecting methods based on convolutional neural networks, have stronger
Robustness can detect more accurate gait cycle in the case where different perspectives, different dressings and belongings.
The object of the present invention is achieved like this:
A kind of gait cycle detecting method based on convolutional neural networks, concrete implementation step are as follows:
Step 1. pre-processes gait video, including video decoding, and pedestrian contour extracts and the normalized figure of mass center
As pretreatment operation；
Step 2. trains the convolutional neural networks for extracting gait cycle feature；
Gait sequence of frames of video to be measured is sent into convolutional neural networks by step 3., and output waveform is led to after filtering
It crosses and determines that the position of adjacent peaks and troughs obtains a gait cycle.
Step 2 specifically:
Step 2.1. indicates position of each video frame in a gait cycle in training set with numerical quantization, and marks
Calculation formula for its label, label value is
Wherein LiThe label value of the i-th frame in gait video, the gait cycle where N indicates the i-th frame includes N frame, n table altogether
Show the i-th frame for n-th frame in the gait cycle at place；
The video frame marked is sent into convolutional neural networks by step 2.2., obtains output valve；
Step 2.3. calculates the error between output valve and label, passes through the more of error back propagation and stochastic gradient descent
Secondary repetitive exercise network is trained until error no longer declines after successive ignition, and the calculation formula of error is
Wherein m is the batch size for inputting network, i.e., every batch of contains m images,For the nerve of corresponding video frame tagging
The estimation of network；
Step 2.4. is saved and is replicated the convolutional neural networks of training completion.
Convolutional neural networks structure described in step 2 includes multilayer convolutional layer and connects the last layer convolutional layer at least
The last layer connection output layer of one layer of full articulamentum, full articulamentum is single neuron.
Step 1 specifically includes:
Step 1.1. carries out sub-frame processing to video sequence, and the sequence after framing is the figure being sequentially arranged
As sequence；
Image sequence comprising pedestrian and background sequence are carried out greyscale transformation by step 1.2., are estimated using median method
The background of entire sequence, and carry out binaryzation and obtain gait contour images,
Dk(x, y)=| Ik(x,y)-Mk(x,y)|
Wherein Ik(x, y) is the gray value at the pixel (x, y) of video sequence kth frame, Mk(x, y) is background ash herein
Angle value, Dk(x, y) is background difference image, and T is selected binarization threshold；
The normalization of step 1.3. profile, carries out unified scaling for all profiles with consistent height, pedestrian contour is returned
One input changed is the content in rectangle tangent with pedestrian contour in each video frame；For interception all in training set
The image of profile traverses its all picture altitude respectively, does ratio as with calibrated altitude；Calibrated altitude under a certain visual angle is
H, K frame image is shared under visual angle, and the height of every frame image sequentially in time is hk, k=1,2 ..., K, then every frame image
Amplification factor be ak=hk/ H applies its corresponding amplification factor a to every frame image is obtained under the visual angle respectivelyk, using double
Linear interpolation algorithm,
fa=f (x, y)+(f (x+1, y)-f (x, y)) × p
fb=f (x, y+1)+(f (x+1, y+1)-f (x, y+1)) × p
In formula, f (x, y) is the gray value at coordinate (x, y) before interpolation, and p and q are weights；Second of linear interpolation is carried out,
Calculating the interpolation result at (x, y) is
G (x, y)=fa+(fb-fa)×q
=(1-p) (1-q) f (x, y)+p (1-p) f (x+1, y)
+q(1-p)f(x,y+1)+pqf(x+1,y+1)
Wherein g (x, y) is the gray value at coordinate (x, y) after interpolation.
The beneficial effects of the present invention are: this method has very strong robust to the variation of angle change, dress ornament and belongings
Property, it solves the problems, such as to be difficult to detect gait cycle under front and back visual angle, existing gait Recognition technology is mostly built
It stands on the basis of the gait cycle of an Accurate Segmentation, the method for the present invention has to Gait Recognition precision in complex environment is improved
Significance can be used for the front end in Gait Recognition, be suitable for security monitoring, human-computer interaction, medical diagnosis and access control system etc.
In identification in.
Detailed description of the invention
Fig. 1 is flow chart of the invention.
Fig. 2 is the effect picture of preceding background separation.
Fig. 3 is a gait cycle and its label value comprising 24 frames.
Fig. 4 is the example of waveform after output waveform and filtering.
Specific embodiment
The present invention will be further described with reference to the accompanying drawing:
Embodiment 1
It is illustrated by taking certain large-scale Gait Recognition database as an example, which includes the gait video sequence of 124 people
Column, everyone 110 sections of videos, including different perspectives, dress ornament and belongings.
Step 1. pre-processes gait video, including, video decoding, pedestrian contour extracts and the figures such as mass center normalization
As pretreatment operation.It first has to carry out sub-frame processing to video sequence, the sequence after framing is to be sequentially arranged
Orderly image sequence, the image sequence comprising pedestrian and background sequence are first subjected to greyscale transformation, substantially not for illumination
For the indoor environment of change, estimate using median method the background of entire sequence, subtract background with foreground image later
Image carries out binaryzation again and obtains gait contour images, and the gray value at the pixel (x, y) of setting video sequence kth frame is Ik(x,
Y), background gray levels herein are Mk(x, y), then background difference image Dk(x, y) and binarization result Bk(x, y) is respectively as follows:
Dk(x, y)=| Ik(x,y)-Mk(x,y)|
Wherein T is selected binarization threshold, and process is as shown in Figure 2.Profile normalization be exactly by all profiles into
Unified scaling go with consistent height, thus when avoiding direction and the distance change of pedestrian and camera, since the depth of field makes to walk
State profile sequence size changes.Four sides of pedestrian contour and frame are made using a rectangular frame for each video frame
It is tangent, using the figure not of uniform size that this frame surrounds as the normalized input of pedestrian contour.It is cut for all in training set
The image of contouring traverses its all picture altitude respectively, does ratio as with calibrated altitude.If the standard under a certain visual angle is high
Degree is H, and K frame image is shared under the angle, and the height of every frame image sequentially in time is hk, k=1,2 ..., K, then often
The amplification factor of frame image
ak=hk/H
To every frame image is obtained under the visual angle, its corresponding amplification factor a is applied respectivelyk, using bilinear interpolation algorithm,
fa=f (x, y)+(f (x+1, y)-f (x, y)) × p
fb=f (x, y+1)+(f (x+1, y+1)-f (x, y+1)) × p
In formula, f (x, y) is the gray value before interpolation at coordinate (x, y), similarly, f (x, y+1) be coordinate before interpolation (x,
Y+1 the gray value at), f (x+1, y+1) are the gray value at coordinate (x+1, y+1) before interpolation, and p and q are weights.Again by faAnd fb
Carry out the interpolation result at second of linear interpolation calculating (x, y):
G (x, y)=fa+(fb-fa)×q
=(1-p) (1-q) f (x, y)+p (1-p) f (x+1, y)
+q(1-p)f(x,y+1)+pqf(x+1,y+1)
Wherein g (x, y) is the gray value at coordinate (x, y) after interpolation.Normalization operation is completed, it is big after being normalized
The small greyscale video frame for being 128 × 88.
Step 2. training convolutional neural networks are for extracting gait cycle feature.
Step 2.1. quantifies each video frame according to its position in gait cycle, and is labeled as its label；
Specifically, choosing the period is 1, the sinusoidal signal that amplitude is 1 is used to characterize the week of gait sequence as low-dimensional signal
Phase property.It being defined in a gait profile sequence, legs joined and right crus of diaphragm have the tendency that video frame advanced in years forward is initial position,
After the sequence of a cycle, the frame that two legs close up again and right crus of diaphragm has trend advanced in years forward is the end position in the period.
This frame image can be used as the end of a cycle at this time, and can be used as the beginning of this period image.When having good positioning
After initial position and final position, the mark value in middle position is calculated according to the SIN function of average value, then has training set
In each frame label value are as follows:
Wherein LiThe label value of the i-th frame in gait video, the gait cycle where N indicates the frame includes N frame, n table altogether
Show that the frame for n-th frame in the gait cycle at place, is illustrated in figure 3 the label value example of a gait cycle comprising 24 frames.
Video frame is sent into convolutional neural networks by step 2.2., obtains output valve；
The video frame marked in training set is sent into convolutional neural networks, the structure of the convolutional neural networks such as can be with
It is: inputs the gray level image for 128 × 88, the combination of first 6 layers respectively 3 convolutional layer and pond layer；First layer is to have 64 5
The convolutional layer that the step-length of × 5 convolution kernels is 1, the second layer is that core is 3 × 3, the pond layer that step-length is 2；Third layer be have 64 3 ×
The convolutional layer that the step-length of 3 convolution kernels is 1, it is 3 × 3 that the 4th layer, which is core, the pond layer that step-length is 2；Layer 5 is to have 64 3 × 3
The convolutional layer that the step-length of convolution kernel is 1, layer 6 is that core is 3 × 3, the pond layer that step-length is 2；7th layer, the 8th layer and 9th layer point
It is not the full articulamentum containing 1024,256 and 1 nodes.
Step 2.3. calculates the error between output valve and label, passes through the more of error back propagation and stochastic gradient descent
Secondary repetitive exercise network；Error is calculated by mean square error (Mean Squared Error, MSE)
Wherein m is the batch size for inputting network, i.e. every batch of contains m image (video frame),For corresponding video frame mark
The estimation of the neural network of label, the i.e. output of neural network.It is trained until error no longer declines after successive ignition.
Step 2.4. is saved and trained convolutional neural networks in copy step 2.3, can be obtained for gait cycle
The convolutional neural networks of property feature extraction and SIN function regression modeling；
Gait sequence of frames of video to be measured is sent into convolutional neural networks by step 3. after the preprocessing process of step 1
In, using frame sequence as horizontal axis, network output is that the longitudinal axis draws waveform, and output waveform passes through the adjacent wave crest of determination after filtering
Or a gait cycle, the waveform of network output as shown in Figure 4 and filtered waveform effect can be obtained in the position of trough,
It is easy to get to its adjacent wave crest and trough, gait cycle can be obtained.
Claims (4)
1. a kind of gait cycle detecting method based on convolutional neural networks, which is characterized in that concrete implementation step are as follows:
Step 1. pre-processes gait video, including video decoding, and pedestrian contour extracts and the normalized image of mass center is pre-
Processing operation；
Step 2. trains the convolutional neural networks for extracting gait cycle feature；
Gait sequence of frames of video to be measured is sent into convolutional neural networks by step 3., and output waveform is after filtering, by true
The position of fixed adjacent peaks and troughs obtains a gait cycle.
2. a kind of gait cycle detecting method based on convolutional neural networks according to claim 1, which is characterized in that step
Rapid 2 specifically:
Step 2.1. indicates position of each video frame in a gait cycle in training set with numerical quantization, and is labeled as it
The calculation formula of label, label value is
Wherein LiThe label value of the i-th frame in gait video, the gait cycle where N indicates the i-th frame includes N frame altogether, and n indicates i-th
Frame is n-th frame in the gait cycle at place；
The video frame marked is sent into convolutional neural networks by step 2.2., obtains output valve；
Step 2.3. calculates the error between output valve and label, is changed by error back propagation and the multiple of stochastic gradient descent
Generation training network, is trained until error no longer declines, the calculation formula of error is after successive ignition
Wherein m is the batch size for inputting network, i.e., every batch of contains m images,For the neural network of corresponding video frame tagging
Estimation；
Step 2.4. is saved and is replicated the convolutional neural networks of training completion.
3. a kind of gait cycle detecting method based on convolutional neural networks according to claim 1, it is characterised in that: step
Convolutional neural networks structure described in rapid 2 includes multilayer convolutional layer and the full connection of at least one layer for connecting the last layer convolutional layer
The last layer connection output layer of layer, full articulamentum is single neuron.
4. a kind of gait cycle detecting method based on convolutional neural networks according to claim 1, it is characterised in that: step
Rapid 1 specifically includes:
Step 1.1. carries out sub-frame processing to video sequence, and the sequence after framing is the image sequence being sequentially arranged
Column；
Image sequence comprising pedestrian and background sequence are carried out greyscale transformation by step 1.2., estimate using median method entire
The background of sequence, and carry out binaryzation and obtain gait contour images,
Dk(x, y)=| Ik(x,y)-Mk(x,y)|
Wherein Ik(x, y) is the gray value at the pixel (x, y) of video sequence kth frame, Mk(x, y) is background gray levels herein,
Dk(x, y) is background difference image, and T is selected binarization threshold；
All profiles are carried out unified scaling with consistent height, pedestrian contour normalization by the normalization of step 1.3. profile
Input be content in each video frame in the rectangle tangent with pedestrian contour；For interception profile all in training set
Image, traverse its all picture altitude respectively, do ratio as with calibrated altitude；Calibrated altitude under a certain visual angle is H, depending on
K frame image is shared under angle, the height of every frame image sequentially in time is hk, k=1,2 ..., K, then every frame image is put
Big multiple is ak=hk/ H applies its corresponding amplification factor a to every frame image is obtained under the visual angle respectivelyk, using bilinearity
Interpolation algorithm,
fa=f (x, y)+(f (x+1, y)-f (x, y)) × p
fb=f (x, y+1)+(f (x+1, y+1)-f (x, y+1)) × p
In formula, f (x, y) is the gray value at coordinate (x, y) before interpolation, and p and q are weights；Second of linear interpolation is carried out, is calculated
Interpolation result at (x, y) is
G (x, y)=fa+(fb-fa)×q
=(1-p) (1-q) f (x, y)+p (1-p) f (x+1, y)
+q(1-p)f(x,y+1)+pqf(x+1,y+1)
Wherein g (x, y) is the gray value at coordinate (x, y) after interpolation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910026947.2A CN109766838B (en) | 2019-01-11 | 2019-01-11 | Gait cycle detection method based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910026947.2A CN109766838B (en) | 2019-01-11 | 2019-01-11 | Gait cycle detection method based on convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109766838A true CN109766838A (en) | 2019-05-17 |
| CN109766838B CN109766838B (en) | 2022-04-12 |
Family
ID=66453724
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910026947.2A Expired - Fee Related CN109766838B (en) | 2019-01-11 | 2019-01-11 | Gait cycle detection method based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109766838B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110598540A (en) * | 2019-08-05 | 2019-12-20 | 华中科技大学 | Method and system for extracting gait contour map in monitoring video |
| CN110765925A (en) * | 2019-10-18 | 2020-02-07 | 河南大学 | Carrier detection and gait recognition method based on improved twin neural network |
| CN110796100A (en) * | 2019-10-31 | 2020-02-14 | 浙江大华技术股份有限公司 | Gait recognition method and device, terminal and storage device |
| CN112329716A (en) * | 2020-11-26 | 2021-02-05 | 重庆能源职业学院 | Pedestrian age group identification method based on gait characteristics |
| CN112989889A (en) * | 2019-12-17 | 2021-06-18 | 中南大学 | Gait recognition method based on posture guidance |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107122707A (en) * | 2017-03-17 | 2017-09-01 | 山东大学 | Video pedestrian based on macroscopic features compact representation recognition methods and system again |
| CN108460340A (en) * | 2018-02-05 | 2018-08-28 | 北京工业大学 | A kind of gait recognition method based on the dense convolutional neural networks of 3D |
| US20180351775A1 (en) * | 2012-12-05 | 2018-12-06 | Origin Wireless, Inc. | Method, apparatus, and system for wireless motion monitoring |
-
2019
- 2019-01-11 CN CN201910026947.2A patent/CN109766838B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180351775A1 (en) * | 2012-12-05 | 2018-12-06 | Origin Wireless, Inc. | Method, apparatus, and system for wireless motion monitoring |
| CN107122707A (en) * | 2017-03-17 | 2017-09-01 | 山东大学 | Video pedestrian based on macroscopic features compact representation recognition methods and system again |
| CN108460340A (en) * | 2018-02-05 | 2018-08-28 | 北京工业大学 | A kind of gait recognition method based on the dense convolutional neural networks of 3D |
Non-Patent Citations (2)
| Title |
|---|
| URIEL MARTINEZ-HERNANDEZ ET AL.: "Recognition of Walking Activity and Prediction of Gait Periods with a CNN and First-Order MC Strate", 《 2018 7TH IEEE INTERNATIONAL CONFERENCE ON BIOMEDICAL ROBOTICS AND BIOMECHATRONICS (BIOROB)》 * |
| 汤荣山 等: "基于卷积神经网络和不完整步态周期的步态识别方法", 《通信技术》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110598540A (en) * | 2019-08-05 | 2019-12-20 | 华中科技大学 | Method and system for extracting gait contour map in monitoring video |
| CN110598540B (en) * | 2019-08-05 | 2021-12-03 | 华中科技大学 | Method and system for extracting gait contour map in monitoring video |
| CN110765925A (en) * | 2019-10-18 | 2020-02-07 | 河南大学 | Carrier detection and gait recognition method based on improved twin neural network |
| CN110765925B (en) * | 2019-10-18 | 2023-05-09 | 河南大学 | Method for detecting carrying object and identifying gait based on improved twin neural network |
| CN110796100A (en) * | 2019-10-31 | 2020-02-14 | 浙江大华技术股份有限公司 | Gait recognition method and device, terminal and storage device |
| CN110796100B (en) * | 2019-10-31 | 2022-06-07 | 浙江大华技术股份有限公司 | Gait recognition method and device, terminal and storage device |
| CN112989889A (en) * | 2019-12-17 | 2021-06-18 | 中南大学 | Gait recognition method based on posture guidance |
| CN112989889B (en) * | 2019-12-17 | 2023-09-12 | 中南大学 | Gait recognition method based on gesture guidance |
| CN112329716A (en) * | 2020-11-26 | 2021-02-05 | 重庆能源职业学院 | Pedestrian age group identification method based on gait characteristics |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109766838B (en) | 2022-04-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109766838A (en) | A kind of gait cycle detecting method based on convolutional neural networks | |
| Niu et al. | Unsupervised saliency detection of rail surface defects using stereoscopic images | |
| CN110084156B (en) | Gait feature extraction method and pedestrian identity recognition method based on gait features | |
| CN105809693B (en) | SAR image registration method based on deep neural network | |
| CN108520216B (en) | Gait image-based identity recognition method | |
| CN101398886B (en) | Rapid three-dimensional face identification method based on bi-eye passiveness stereo vision | |
| CN108537191B (en) | Three-dimensional face recognition method based on structured light camera | |
| CN104063702B (en) | Three-dimensional gait recognition based on shielding recovery and partial similarity matching | |
| CN109409190A (en) | Pedestrian detection method based on histogram of gradients and Canny edge detector | |
| CN105277567B (en) | A kind of fabric defects detection method | |
| CN108764186A (en) | Personage based on rotation deep learning blocks profile testing method | |
| CN107451609A (en) | Lung neoplasm image identification system based on depth convolutional neural networks | |
| CN106875395A (en) | Super-pixel level SAR image change detection based on deep neural network | |
| CN107967695A (en) | A kind of moving target detecting method based on depth light stream and morphological method | |
| CN107633272B (en) | DCNN texture defect identification method based on compressed sensing under small sample | |
| CN103971329A (en) | Cellular nerve network with genetic algorithm (GACNN)-based multisource image fusion method | |
| Chunli et al. | A behavior classification based on enhanced gait energy image | |
| CN102880870B (en) | The extracting method of face characteristic and system | |
| Xie et al. | Fabric defect detection method combing image pyramid and direction template | |
| Han et al. | An improved corner detection algorithm based on harris | |
| CN111858997B (en) | Cross-domain matching-based clothing template generation method | |
| CN105550703A (en) | Image similarity calculating method suitable for human body re-recognition | |
| CN103824057A (en) | Pig respiratory rate detection method based on area feature operator | |
| CN109544575A (en) | One kind being based on the matched method for reconstructing 3 D contour of ISAR sequence polygon | |
| CN104574400A (en) | Remote sensing image segmenting method based on local difference box dimension algorithm |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20220412 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |