CN106901741A - A kind of respiratory rate detection method suitable for environment round the clock - Google Patents
A kind of respiratory rate detection method suitable for environment round the clock Download PDFInfo
- Publication number
- CN106901741A CN106901741A CN201710145390.5A CN201710145390A CN106901741A CN 106901741 A CN106901741 A CN 106901741A CN 201710145390 A CN201710145390 A CN 201710145390A CN 106901741 A CN106901741 A CN 106901741A
- Authority
- CN
- China
- Prior art keywords
- video
- respiratory
- subband
- chest
- respiratory rate
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/08—Detecting, measuring or recording devices for evaluating the respiratory organs
- A61B5/0816—Measuring devices for examining respiratory frequency
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
- A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
- A61B5/113—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb occurring during breathing
-
- 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
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Surgery (AREA)
- Biomedical Technology (AREA)
- Veterinary Medicine (AREA)
- Public Health (AREA)
- Pulmonology (AREA)
- Physiology (AREA)
- Animal Behavior & Ethology (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- General Physics & Mathematics (AREA)
- Human Computer Interaction (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Dentistry (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention discloses a kind of respiratory rate detection method suitable for environment round the clock, its implementation process is：（1）Control camera gathers 60s just to the video of face, and ensures that upper half of human body video is in collection picture；（2）Video to gathering carries out Face datection and chest positioning action, the chest sport video stablized；（3）Color space conversion is carried out to video；（4）Respiratory movement sequence is extracted from the video of chest using video motion amplifying technique；（5）Maximal possibility estimation is carried out to respiratory movement sequence, and the optimization of respiratory waveform is carried out using the first estimation frequency combination smothing filtering being calculated, finally carry out estimating again for respiratory rate using peak point detection technique.The present invention determines the main region that breathing occurs by Face datection and chest location technology, and the stability and accuracy of respiratory rate detection are improve in the respiratory movement sequence that can stablize respectively under environment round the clock by the video motion amplifying technique based on phase.
Description
Technical field
The present invention relates to medical video image processing field, specifically a kind of respiratory rate detection side suitable for environment round the clock
Method.
Background technology
With the development of computer technology, video image processing technology has obtained rapidly development and has applied in recent years, and
And the every aspect lived has been penetrated into, it is that people bring many facilities.In medical science field, video processing technique is in doctor
Treat diagnosis, surgical guidance and routine testing aspect have huge application prospect, for the development of modern medicine provide it is positive
Auxiliary support is acted on.
Respiratory rate is one of sensitive indicator of acute respiration dysfunction disease detection, and to long-term respiratory rate variation tendency
Monitoring analysis can also reflect the health status of human body, such as physical fatigue state, sleep quality.So, respiratory rate detection
It is significant in terms of relevant disease prevention and routine testing.Conventional respiratory rate detection method is b12extrocardiography
(Electrocardiogram, ECG), i.e., estimate respiratory rate indirectly by electrocardiogram (ECG) data.Although the method accuracy is higher,
Gained breathing rate score is not instant respiratory rate, but is estimated to obtain indirectly by electrocardiogram (ECG) data, and is needed in quilt during measurement
Adhesive electrode is measured with survey person, can bring body and psychological discomfort, influence respiratory rate measurement knot to testee
Really.Therefore, the appearance of contactless respiratory rate detection technique can solve the above problems well.But the application heat for proposing in the early time
The non-contact detection modes such as image camera, miniature organism radar, 3D body-sensings camera, magnetic induction phase-shifting technique because it is costly,
The reason such as equipment volume is big, anti-interference is poor and routine testing cannot be applied to, so inexpensively, convenient, efficient noncontact
The detection of formula respiratory rate is still the emphasis direction studied from now on.
According to medical definition, a secondary fluctuation of chest is defined as time air-breathing of respiration, i.e., one and the process once exhaled,
So the number of times for counting chest rise per minute can estimate respiratory rate.Contactless respiratory rate detection skills most of at present
Art is all to estimate respiratory rate by counting the number of times of chest rise motion in a minute.Alinovi et al. is proposed in the recent period
Contactless video respiratory rate detection technique based on spatial temporal dissolution.The method uses EVM (Eulerian Video
Magnification) thought, first by Laplacian pyramid raw video image, extracts different scale hypograph
Brightness information；Then breath signal is extracted and amplified using space-time processing techniques；Finally use maximum likelihood estimate
Signal Analysis estimate respiratory rate.
Although Alinovi methods can exactly estimate respiratory rate, there is following defect：(1) put based on EVM motions
Big mode can also carry out the amplification of equal magnitude while prominent respiratory movement to the noise in video, and noise meeting after amplifying
Extraction to later stage breath signal produces interference, influences the estimation of respiratory rate；(2) former method uses Global treatment mode, the party
Although method can farthest obtain respiratory movement signal, other disturbed motions can be produced to the judgement of breath signal in scene
Raw interference；(3) require to carry out the scene of video acquisition at night for sleep-respiratory detection etc., common camera cannot be effectively
Collection video.
The content of the invention
It is an object of the invention to provide a kind of respiratory rate detection method suitable for environment round the clock, to solve prior art
Alinovi methods estimate the problem that respiratory rate is present.
In order to achieve the above object, the technical solution adopted in the present invention is：
A kind of respiratory rate detection method suitable for environment round the clock, it is characterised in that：Comprise the following steps：
(1) camera, is used to shoot human body to obtain video, the camera is soft by Matlab by computer
Part is controlled, and ensures that upper half of human body is in video pictures and as far as possible in central area during shooting, and keep camera with solid
Fixed resolution ratio, frame per second and RGB color space shoot certain hour, there is shooting under stable incandescent light source environment daytime in room
Video saves as AVI format, and the video shot under night no light environment saves as mp4 forms；
(2), the video to camera collection carries out Face datection, and according to human geometry's knowledge, is positioned by human face region
Chest position, obtains the chest sport video of stabilization, and process is as follows：
(2.1) human face region, is obtained：
In a computer using the first frame of the video of camera collection as reference picture, by the Viola- in OpenCV
Jones human-face detectors detect the human face region of rectangle, obtain four coordinates of summit F1 of rectangle frame；
(2.2) breath signal pickup area, is determined：
According to priori, the height of people can be described with head height, cause chest region of variation substantially to exist by breathing
Known according to priori to the region above the 6th head height line the lower zone of the 6th head height line and the 7th head height line, i.e. shoulder lower section
Knowledge can find out breathing from video and cause chest region of variation, then determine breath signal in breathing causes chest region of variation
Pickup area, it is determined that principle it is as follows：
Selection since the crown, move down 1.4-1.6 times of head height as breath signal pickup area starting ordinate, with
Face frame central point is moved to the left the 0.23-0.27 times of head breadth as breath signal pickup area starting abscissa, 0.4-0.6 times
Head height and 0.8-1.2 times of the head breadth respectively as breath signal pickup area length with it is wide；
(2.3) chest sport video, is obtained：
After determining breath signal pickup area, video is cut to retain the corresponding chest of breath signal pickup area
Location drawing picture, and export and carry out after the location drawing picture sequence of chest video preservation, generate the chest sport video of stabilization；
(3) color space of chest sport video, is converted into Gray-Scale by RGB, to obtain Gray-Scale colors
The video of color space, it is simultaneously effective by the protrusion that monochrome information in triple channel exercise information integration to single channel, will be realized
Data computation complexity is reduced, RGB and Gray-Scale conversion formulas are as follows：
(4), the video to Gray-Scale color spaces carries out motion amplification, to extract the respiratory movement signal of stabilization,
Detailed process is as follows：
(4.1), space phase is decomposed：
Space phase treatment is carried out to image using plural steerable pyramid, by iterate to calculate by picture breakdown into
The subband sequence of different scale, phase directional, the motion phase information of image is just lain in subband sequence, and plural direction is controllable
Pyramid implementation process is as follows：
(4.1.1), input image size, determine Scale Decomposition number of plies M：
M=floor (log2Min (h, w)) -2 (2),
In formula, h and w represents the height and width of original image, and floor is represented and rounded downwards；
(4.1.2), the filtering directioin parameter N (according to experiment effect, this default setting N=4) for determining wave filter；
(4.1.3), input picture are first broken down into high pass subband H0With low pass subband L0, wherein low pass subband comprising figure
As global information, high pass subband includes image detail, during wherein phase information lies in low pass subband；
(4.1.4), according to parameter N to L0The controllable filtering of travel direction, obtains the subband sequence B of N class different directionsk(k=
0,1 ... N) and subband part L1；
(4.1.5), by subband part L1After carrying out two sampling, (4.1.4) operation is repeated, until the Scale Decomposition number of plies reaches
M；
After the completion of (4.1.6), operation, M*N+2 (H are obtained altogether0With LM+1) individual subband sequence, wherein M*N subband sequence B
In include respirometric phase information, for follow-up signal treatment；
(4.2), time bandpass filtering：
Subband sequence is temporally filtered using preferable bandpass filter, passband is 0.05~1.25Hz；
(4.3), signal amplifies：
For filtered subband sequence, the amplification that default amplification factor α is moved directly is multiplied by；
(4.4) breath signal is extracted, and process is as follows：
(4.4.1), two-value conversion is carried out to the subband sequence after amplification, by default threshold gammathProminent respiratory movement,
Formula is as follows：
Wherein l is represented through the l straton bands after the controllable pyramid decomposition of direction；X, y represent the corresponding image of respective sub-bands
Size, i represents the frame number of video；α represents amplification factor；FlRepresent the filtered sequential value of corresponding subband, BlRepresent correspondence
Sequential value with thresholding；
(4.4.2), image sequence is changed into average phase signal to characterize body kinematics pattern, i.e. respiratory movement.Averagely
Phase signal formula is as follows：
By behind formula (3), (4), will obtain characterizing respirometric multi channel signals, each passage is entered into row of channels average
After operation, average respiratory waveform can be obtained, that is, the respiratory movement signal stablized；
(5) frequency-domain analysis, is carried out to the multichannel respiratory waveform for obtaining, respiratory rate is carried out by maximum likelihood estimate
According to a preliminary estimate, coordinate smothing filtering that optimization is filtered to average respiratory waveform after obtaining initial respiratory rate, obtain accurately
Respiratory waveform, and respiratory rate is carried out using peak point detection method estimate that detailed process is as follows again：
(5.1) maximal possibility estimation：
The L obtained by formula (4)lRepresent the breath signal trend of different scale and phase directional, different LlTo same fortune
The different descriptions of animal body same object is shot equivalent to the video camera of diverse location and angle produced by multigroup figure
As result, respiratory rate estimation is carried out using maximum likelihood method based on this thought, by the signal of be polymerized multiple yardsticks, phase directional
Sequence reaches the effect of enhancing breath signal, and specific formula is as follows：
Wherein, fsThe sample frequency of signal is represented, N represents collection video totalframes, and M represents each chi after pyramid decomposition
The total quantity of degree and directional subband, DFT { } represents discrete Fourier transform, and argmax operations represent and energy is obtained from set
Value maximum point index；
(5.2), smothing filtering and peak point are detected：
With f0On the basis of, passband is respectively widened into 0.05H up and downzSmothing filtering is carried out to average respiratory waveform, filtering is obtained
Respiratory waveform after optimization, now clearly, the fluctuating of each of which group waveform represents one to the respiratory waveform of respiratory waveform
Secondary breathing rises and falls, so the crest quantity counted in this section of waveform can obtain respiratory rate, therefore the waveform is used
Peak point sense command in Matlab tool boxes.
A kind of described respiratory rate detection method suitable for environment round the clock, it is characterised in that：In step (1), on daytime
Interior has under stable incandescent light source environment, and video is held with the resolution ratio of 640*480, the frame per second and RGB color space of 30 frames/second
Continuous collection 60 seconds, and preserve into AVI format；Under night no light environment, video is with the resolution ratio of 1280*720, the frame of 9 frames/second
Rate and RGB color space continuous collecting 60 seconds, and preserve into mp4 forms.
A kind of described respiratory rate detection method suitable for environment round the clock, it is characterised in that：In step (2.1), at night
Between in the case of, there is face flase drop with video quality reason because night scenes noise is larger in Viola-Jones human-face detectors
Situation, can using repair algorithm be processed.
The present invention is only imaged by with common without using electrode or sensor contacts human body with night vision function
Head can automatically monitor the important physical signs of human body, respiratory rate (Respiratory Rate).The method of proposition is by people
Body chest video is acquired and respective handling, can effectively position breathing main region, eliminates the interference of surrounding scene noise, from
The accurate breath signal that extracts is estimated for respiratory rate in chest rise.At daytime or indoor stabilization incandescent light source situation and night
In the case of two kinds of no light, the respiratory rate that the present invention estimates has good result uniformity with true respiratory rate, stabilization
Property is higher.
The present invention has the following advantages that compared with prior art：
1) present invention has that complex operation is fettered with limbs for conventional contact respiratory rate detection technique, proposes
A kind of contactless respiratory rate detection technique based on chest Video processing.The technology is without using electrode or special sensor
Human body is contacted to obtain signal, and need to only be breathed by automatic detection by the common camera with night vision function
Rate.Equipment cost is significantly reduced, versatility is improve, and improve the level of comfort of tester, it is adaptable to daily to exhale
Inhale monitoring and relevant disease prevention.
2) present invention is changed by considering the chest area that breathing causes, and effectively selects breath signal to extract region, gram
The interference of other noises in scene is taken, and has suitably reduced the time complexity of algorithm；Using the motion amplification skill based on phase
Art, extracts respiratory movement sequence and it is amplified, and obtains the respiratory movement signal of stabilization；Night is realized using infrared camera
Between video acquisition, and by phase information extracting mode, respiratory rate can be exactly detected under environment round the clock.
3) present invention carries out maximal possibility estimation analysis by breath signal, realizes the preliminary of breath signal frequency and estimates
Meter；Then using just estimating that frequency carries out smothing filtering to breath signal, optimize respiratory waveform, make respiratory movement trend brighter
It is aobvious, respiratory rate estimation is carried out to waveform after optimization finally by peak value point detecting method, improve the precision of respiratory rate detection.
Brief description of the drawings
Fig. 1 is the FB(flow block) of contactless respiratory rate detection of the invention.
Fig. 2 is equipment installation effect figure of the invention.
Fig. 3 is Face datection design sketch (daytime).
Fig. 4 is Face datection design sketch (night flase drop situation).
Fig. 5 is Face datection design sketch (situation after the reparation of night flase drop).
Fig. 6 is that chest positions schematic diagram (daytime).
Fig. 7 is that chest positions schematic diagram (night).
Fig. 8 is the respiratory movement waveform that the present invention is extracted.
Fig. 9 is respiratory movement waveform after smothing filtering of the present invention.
Specific embodiment
Reference picture 1, the implementation steps of respiratory rate detection of the present invention are as follows：
Be positioned over the IP Camera that has night vision function just to face front about 0.5m by step 1, reference picture 2
Position, it is also possible to frame is attached between camera and computer in the top of computer display using USB.Operator
Video is shot using Matlab softwares control camera, it is ensured that upper half of human body is completely within video pictures, and is located as far as possible
In central area.In the case where having stable incandescent light source environment in room daytime, video is with the resolution ratio of 640*480, the frame of 30 frames/second
Rate and RGB color space continuous collecting 60 seconds, and preserve into AVI format.Under night no light environment, due to camera hardware limit
Reason processed, video with the resolution ratio of 1280*720, the frame per second of 9 frames/second and RGB color space continuous collecting 60 seconds.
Step 2, the video to gathering carries out Face datection, and according to human geometry's knowledge, chest is positioned by human face region
Position, obtains the chest sport video of stabilization, has 2 important improvements compared to prior art.
2a) using the first frame of input video as reference picture, examined by the Viola-Jones faces in OpenCV first
The human face region that device detects rectangle is surveyed, four coordinates of summit F1 of rectangle frame is obtained, under daylight environment as shown in Figure 3.For
Under nighttime conditions, Viola-Jones can be because there are the feelings of face flase drop in the larger reason such as with video quality of night scenes noise
Condition, the present invention is processed using algorithm is repaired, and solves the problems, such as most of flase drop, repairs algorithm principle：Due to being adopted in video
Face is the main target of scene during collection, so occurring the wrong block diagram width respectively less than face of flase drop as a rule
Block diagram width.Therefore when multiple possible " face " block diagrams of VJ detector acquisitions, can give tacit consent to and retain the maximum coordinates regional of width,
As face block diagram.Specific Fig. 4, shown in 5, when there is flase drop situation (Fig. 4) in original VJ detectors, flase drop " face " width of frame
It is smaller than correct face width of frame, after treatment, only retain correct face (Fig. 5).
2b) height of people can be described with head height.According to priori, aduit height is usually 8 times of head height, from
Sole rises upwards, and the chest of male is about above the 6th head height, and women is then lower slightly with respect to male.So, chest is caused by breathing
Portion's regional change is substantially in the 6th head height line and the lower zone of the 7th head height line, that is, shoulder lower section is on the 6th head height line
The region of side.Binding experiment effect, have selected a part as breath signal pickup area from above-mentioned zone herein.Selection area
Domain is：Since the crown, 1.5 times of head heights are moved down as breathing area starting ordinate, with face frame central point to moving to left
0.25 times of head breadth is moved as breathing area starting abscissa, the length of 0.5 times of head height and 1 times of the head breadth respectively as pickup area
With width, specific such as Fig. 6,7.
After 2c) being correctly detecting chest position, subsequent video is cut, output image sequence is according to step 1 basis
Different usage scenarios carries out video preservation, chest sport video of surviving.
Operation above has 2 points of improvement compared to prior art：1. the area of generation is clearly breathed by chest location technology
Domain, significantly reduces the time complexity of algorithm, and after determining chest region, can remove making an uproar for non-respiratory region in scene
Acoustic jamming, it is to avoid because of the estimated bias problem that overall situation estimation is produced by noise problem；2. face flase drop reparation can be most
Accuracy of detection of the tradition Viola-Jones under nighttime conditions is lifted in the case of number, chest position success rate is lifted, increases breathing
The stability of signal extraction.
Step 3, the color space of chest video is converted to Gray-Scale by RGB, and triple channel exercise information integration is arrived
On single channel, the protrusion for realizing monochrome information effectively reduces data computation complexity simultaneously.RGB and Gray-Scale changes public
Formula is as follows：
Gray=0.2989*R+0.5870*G+0.1140*B (1),
Step 4, the video to Gray-Scale color spaces carries out motion amplification, extracts respiratory movement signal.
4a) space phase is decomposed
Space phase treatment is carried out to image using plural steerable pyramid, by iterate to calculate by picture breakdown into
The subband sequence of different scale, phase directional, the motion phase information of image is just lain in subband sequence.With traditional pyramid
Decompose different, plural steerable pyramid, neatly can into multiple directions by picture breakdown by direction initialization phase angle
The sub-band information of control, and all directions subband have without aliasing, quadrature in phase the characteristics of.Plural steerable pyramid was realized
Cheng Wei：
(1) input image size, determines Scale Decomposition number of plies M
M=floor (log2Min (h, w)) -2 (2),
In formula, h and w represents the height and width of original image, and floor is represented and rounded downwards.
(2) the phase directional parameter N of wave filter is determined
(3) input picture is first broken down into H0With low pass subband L0, wherein low pass subband includes image global information, high
Logical subband includes image detail, during wherein phase information lies in low pass subband.
(4) according to parameter N to L0The controllable filtering of travel direction, obtains the subband sequence B of N class different directionsk(k=0,1 ...
) and subband part L N1。
(5) L is entered into subband part1After row two is sampled, (4) operation is repeated, until the Scale Decomposition number of plies reaches M
(6) after the completion of operating, M*N+2 (H are obtained altogether0With LM+1) individual subband sequence.Wherein, in (M*N) individual subband sequence B
Comprising respirometric phase information, for follow-up signal treatment.
4b) time bandpass filtering
Subband sequence is temporally filtered using preferable bandpass filter, passband is 0.05~1.25Hz (correspondences
3~75 beats/min of respiratory rate).
4c) signal amplifies
For filtered subband sequence, the amplification that amplification factor α is moved directly is multiplied by, acquiescence amplification factor is 10
Times.
4d) breath signal is extracted
(1) threshold transition is carried out to subband sequence after amplification, by threshold gammath(according to experiment effect, this is set as
10) prominent respiratory movement, process is as follows：
Wherein l is represented through the l straton bands after the controllable pyramid decomposition of direction；X, y represent the corresponding image of respective sub-bands
Size, i represents the frame number of video；α represents amplification factor, and acquiescence amplification factor multiple is 10；FlRepresent corresponding subband filtering
Sequential value afterwards, BlRepresent the sequential value of corresponding subband thresholding.
(2) image sequence conversion average phase signal is characterized into body kinematics pattern, i.e. respiratory movement.Average phase is believed
Number formula is as follows：
By formula (3), (4), the present invention obtains characterizing respirometric multi channel signals, each passage is entered into row of channels and is put down
After operating, average respiratory waveform can be obtained, see Fig. 8.
Step 5, the multichannel respiratory waveform to obtaining carries out frequency-domain analysis, and respiratory rate is carried out by maximum likelihood estimate
According to a preliminary estimate, cooperation smothing filtering carries out smothing filtering, optimization breathing to average respiratory waveform after obtaining initial respiratory rate
Waveform, and respiratory rate is carried out using peak point detection method estimate again.
5a) maximal possibility estimation
The L obtained by formula (4)lRepresent the breath signal trend of different scale and phase directional.Different LlTo same fortune
The different descriptions of animal body same object is shot equivalent to the camera of diverse location and angle produced by multigroup figure
As result.Based on this thought, the present invention carries out respiratory rate estimation using maximum likelihood method, by multiple yardsticks, the phase side of being polymerized
To signal sequence reach enhancing breath signal effect.Specific formula is as follows：
Wherein, fsThe sample frequency of signal is represented, N represents collection video totalframes；M represents each chi after pyramid decomposition
The total quantity of degree and directional subband；DFT { } represents discrete Fourier transform.
5b) smothing filtering is detected with peak point
With f0On the basis of, passband is respectively widened 0.05Hz up and down carries out smothing filtering to average respiratory waveform, obtains filter
Respiratory waveform after ripple optimization, is shown in Fig. 9.Now, the respiratory waveform of respiratory waveform is clearly.Matlab is used to the waveform
Peak point sense command in tool box, it is specific as follows：
Data=detrend (data)；
[pks,~]=findpeaks (data, ' minpeakheight', 0, ' minpeakdistance', 14)；
M=size (pks)；
Output=round (m (2) * (60*Fs)/len)；
Wherein, data represents smothing filtering postamble sequence；Function detrend () represents that signal goes trending to process；
Function findpeaks () represents that the parameter logistics such as minimum peak height, the peak-to-peak distance according to setting evidence carries out crest statistics；
M represents crest quantity；Fs represents sample frequency；Len represents signal length；Function round () represents and estimated result is carried out
Round；Output estimates respiratory rate for last.
So far, it is adaptable to which the respiratory rate detection technique of environment is basically completed round the clock.
Effectiveness of the invention is further illustrated below by way of the respiratory rate test experience under daylight environment and nighttime conditions.
First, daylight environment respiratory rate test experience
1. video parameter：
Video parameter such as table 1：
Video parameter under the daylight environment of table 1
2. experiment content：
It is checking accuracy of the invention, this experiment is carried out in the range of 6-36 beats/min of respiratory rate to tester
Multiple video acquisition and respiratory rate detection.Tester carries out exhaling for the specific frequency with eupnea or according to display screen prompting respectively
Suction is referred to as true respiratory rate.
In order to carry out quantitative assessment to experimental result, this carries out Performance Evaluation using 3 kinds of evaluation indexes.1st index
It is mean error Me, i.e. the deviation of measured value and actual value is shown in formula (6):
2nd index is root-mean-square error, is denoted as RMSE (Root Mean Square Error), and RMSE is represented closer to 0
Respiratory rate detection technique is more stable, and robustness more preferably, is shown in formula (7):
3rd index is the Average Accuracy RR of respiratory rateac, see formula (8)：
In formula, N represents video total quantity, RRevRepresent respiratory rate estimate, RRtrueRepresent actual respiratory rate value.
It is checking inventive energy, the contactless respiratory rate detection method that the present invention has reappeared Alinovi is tested
Contrast.Respiratory rate detection performance as shown in table 2 in the case of daytime
Respiratory rate detection performance comparison under the daylight environment of table 2
3. respiratory rate Analysis of test results：
Hair is stablized under the daylight environment of incandescent light source indoors, and two kinds of technologies are capable of achieving the accurate detection of respiratory rate.
But because Alinovi methods are easily disturbed and noise problem when the overall situation is estimated by scene, so in later stage power Spectral Estimation
When can produce energy leakage effect, cause respiratory rate estimate there is relatively large deviation.And the present invention passes through chest positioning and optimizing and exhales
Inhale region and signal stabilization is ensured by PVM outstanding noiseproof feature, therefore had a distinct increment in experimental result.So
In table 2, the present invention has lifting in performance parameter.
2nd, night-environment respiratory rate test experience
1. video parameter：
Video parameter such as table 3：
Video parameter under the daylight environment of table 3
2. experiment content：
It is checking accuracy of the invention, this experiment is carried out in the range of 9-23 beats/min of respiratory rate to tester
Multiple video acquisition and respiratory rate detection.Tester carries out exhaling for the specific frequency with eupnea or according to display screen prompting respectively
Suction is referred to as true respiratory rate.
Same three kinds of indexs using in experiment one of the invention are evaluated respiratory rate detection technique, the method for contrast
It is identical.Respiratory rate detection performance under night-environment is as shown in table 4.
Respiratory rate detection performance comparison under the night-environment of table 2
3. respiratory rate Analysis of test results：
Found in contrast：In the environment of no light in night room, by using infrared camera, the present invention realizes night
Between under environment respiratory rate detection.But under the conditions of identical device, the present invention is obviously improved compared to Alinovi.It is main former
Because being when measured is in camera closer distance, the infrared light major part that camera sends all is reflected by human body, so
Measured's display effect in video is higher than the brightness value of scene around, is reflected in greyscale video i.e. " partially white ".Institute
Monochrome information cannot effectively be extracted with algorithm of the tradition based on brightness value change detection signal, so respiratory rate detection occurs seriously
Deviation, it is impossible to suitable for the detection of night respiration rate.And context of methods is changed the extraction of signal from phase angle, it is prevented effectively from
Brightness value extracts difficult defect in conventional method, it is possible to achieve relatively stable respiratory rate detection.But result is still relatively white
It occurs declining, and its reason is：Current night camera head is generally security device, and video is carried out using high compression ratio technology
During compression can to the aspects such as the definition of video, image detail produce loss, and 9 frames/second frame per second compared to 30 frames on daytime/second
Frame per second have larger downslide, reduce data volume, breath signal is extracted and produces certain influence.But from MeWith RRacTwo parameter can
To find out, even if under nighttime conditions, the present invention still can keep rate of coincideing higher with true respiratory rate, meet daily making
With requiring.
Claims (3)
1. a kind of respiratory rate detection method suitable for environment round the clock, it is characterised in that：Comprise the following steps：
(1) camera, is used to shoot human body to obtain video, the camera passes through Matlab software controls by computer
System, ensures that upper half of human body is in video pictures and as far as possible in central area, and keep camera with fixed during shooting
Resolution ratio, frame per second and RGB color space shoot certain hour, there is the video shot under stable incandescent light source environment daytime in room
AVI format is saved as, the video shot under night no light environment saves as mp4 forms；
(2), the video to camera collection carries out Face datection, and according to human geometry's knowledge, chest is positioned by human face region
Position, obtains the chest sport video of stabilization, and process is as follows：
(2.1) human face region, is obtained：
In a computer using the first frame of the video of camera collection as reference picture, by the Viola- in OpenCV
Jones human-face detectors detect the human face region of rectangle, obtain four coordinates of summit F1 of rectangle frame；
(2.2) breath signal pickup area, is determined：
According to priori, the height of people can be described with head height, and chest region of variation is caused substantially the 6th by breathing
The lower zone of head height line and the 7th head height line, i.e. shoulder lower section, can according to priori to the region above the 6th head height line
Breathing is found out from video and causes chest region of variation, breath signal collection is then determined in breathing causes chest region of variation
Region, it is determined that principle it is as follows：
Selection moves down 1.4-1.6 times of head height as breath signal pickup area starting ordinate, with face since the crown
Frame central point is moved to the left the 0.23-0.27 times of head breadth as breath signal pickup area starting abscissa, 0.4-0.6 times of head height
The head breadth with 0.8-1.2 times respectively as breath signal pickup area length with it is wide；
(2.3) chest sport video, is obtained：
After determining breath signal pickup area, video is cut to retain the corresponding chest position of breath signal pickup area
Image, and export and carry out after the location drawing picture sequence of chest video preservation, generate the chest sport video of stabilization；
(3) color space of chest sport video, is converted into Gray-Scale by RGB, it is empty to obtain Gray-Scale colors
Between video, by triple channel exercise information integration to single channel, will realize monochrome information protrusion simultaneously effectively reduce
Data computation complexity, RGB and Gray-Scale conversion formulas are as follows：
(4), the video to Gray-Scale color spaces carries out motion amplification, to extract the respiratory movement signal of stabilization, specifically
Process is as follows：
(4.1), space phase is decomposed：
Space phase treatment is carried out to image using plural steerable pyramid, by iterating to calculate picture breakdown into difference
The subband sequence of yardstick, phase directional, the motion phase information of image is just lain in subband sequence, the plural controllable golden word in direction
Tower implementation process is as follows：
(4.1.1), input image size, determine Scale Decomposition number of plies M:
M=floor (log2min(h,w))-2 (2)
In formula, h and w represents the height and width of original image, and floor is represented and rounded downwards；
(4.1.2), the filtering directioin parameter N (according to experiment effect, this default setting N=4) for determining wave filter；
(4.1.3), input picture are first broken down into high pass subband H0With low pass subband L0, wherein low pass subband comprising image it is complete
Office's information, high pass subband includes image detail, during wherein phase information lies in low pass subband；
(4.1.4), according to parameter N to L0The controllable filtering of travel direction, obtains the subband sequence B of N class different directionsk(k=0,
1 ... N) and subband part L1；
(4.1.5), by subband part L1After carrying out two sampling, (4.1.4) operation is repeated, until the Scale Decomposition number of plies reaches M；
After the completion of (4.1.6), operation, M*N+2 (H are obtained altogether0With LM+1) individual subband sequence, included in wherein M*N subband sequence B
Respirometric phase information, for follow-up signal treatment；
(4.2), time bandpass filtering：
Subband sequence is temporally filtered using preferable bandpass filter, passband is 0.05~1.25Hz；
(4.3), signal amplifies：
For filtered subband sequence, the amplification that default amplification factor α is moved directly is multiplied by；
(4.4) breath signal is extracted, and process is as follows：
(4.4.1), two-value conversion is carried out to the subband sequence after amplification, by default threshold gammathProminent respiratory movement, formula
It is as follows：
Wherein l is represented through the l straton bands after the controllable pyramid decomposition of direction；X, y represent the corresponding picture size of respective sub-bands,
I represents the frame number of video；α represents amplification factor；FlRepresent the filtered sequential value of corresponding subband, BlRepresent corresponding subband threshold
The sequential value of value；
(4.4.2), image sequence is changed into average phase signal to characterize body kinematics pattern, i.e. respiratory movement.
Average phase signal formula is as follows：
By behind formula (3), (4), will obtain characterizing respirometric multi channel signals, each passage is entered into row of channels average operation
Afterwards, average respiratory waveform can be obtained, that is, the respiratory movement signal stablized；
(5) frequency-domain analysis, is carried out to the multichannel respiratory waveform for obtaining, the preliminary of respiratory rate is carried out by maximum likelihood estimate
Estimate, coordinate smothing filtering that optimization is filtered to average respiratory waveform after obtaining initial respiratory rate, acquisition is accurately breathed
Waveform, and respiratory rate is carried out using peak point detection method estimate that detailed process is as follows again：
(5.1) maximal possibility estimation：
The L obtained by formula (4)lRepresent the breath signal trend of different scale and phase directional, different LlTo same moving object
The different descriptions of body same object is shot equivalent to the video camera of diverse location and angle produced by multiple series of images knot
Really, respiratory rate estimation is carried out using maximum likelihood method based on this thought, by the signal sequence of be polymerized multiple yardsticks, phase directional
The effect of enhancing breath signal is reached, specific formula is as follows：
Wherein, fsRepresent the sample frequency of signal, N represents collection video totalframes, M represent after pyramid decomposition each yardstick and
The total quantity of directional subband, DFT { } represents discrete Fourier transform, and argmax operations represent and energy value maximum is obtained from set
Point index；
(5.2), smothing filtering and peak point are detected：
With f0On the basis of, passband is respectively widened into 0.05H up and downzSmothing filtering is carried out to average respiratory waveform, filtering optimization is obtained
Respiratory waveform afterwards, now clearly, the fluctuating of each of which group waveform is represented once exhales the respiratory waveform of respiratory waveform
Volt is picked up, so the crest quantity counted in this section of waveform can obtain respiratory rate, therefore Matlab works is used to the waveform
Peak point sense command in tool case.
2. a kind of respiratory rate detection method suitable for environment round the clock according to claim 1, it is characterised in that：Step
(1) in, in the case where having stable incandescent light source environment in room daytime, video with the resolution ratio of 640*480, the frame per second of 30 frames/second and
RGB color space continuous collecting 60 seconds, and preserve into AVI format；Under night no light environment, video is with the resolution of 1280*720
Rate, the frame per second of 9 frames/second and RGB color space continuous collecting 60 seconds, and preserve into mp4 forms.
3. a kind of respiratory rate detection method suitable for environment round the clock according to claim 1, it is characterised in that：Step
(2.1) in, under nighttime conditions, Viola-Jones human-face detectors are because night scenes noise is larger with video quality reason
There is the situation of face flase drop, can be processed using algorithm is repaired.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710145390.5A CN106901741A (en) | 2017-03-13 | 2017-03-13 | A kind of respiratory rate detection method suitable for environment round the clock |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710145390.5A CN106901741A (en) | 2017-03-13 | 2017-03-13 | A kind of respiratory rate detection method suitable for environment round the clock |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106901741A true CN106901741A (en) | 2017-06-30 |
Family
ID=59187791
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710145390.5A Pending CN106901741A (en) | 2017-03-13 | 2017-03-13 | A kind of respiratory rate detection method suitable for environment round the clock |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106901741A (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107550509A (en) * | 2017-10-16 | 2018-01-09 | 合肥工业大学 | A kind of X-ray machine imaging triggering method based on the detection of contactless breathing pattern |
| CN107967684A (en) * | 2017-12-11 | 2018-04-27 | 深圳大学 | Contactless sleep-respiratory detection method and device |
| CN109063763A (en) * | 2018-07-26 | 2018-12-21 | 合肥工业大学 | Video minor change amplification method based on PCA |
| CN109124613A (en) * | 2018-08-17 | 2019-01-04 | 西安电子科技大学 | Contactless heart rate monitoring unit and method based on interactive voice |
| CN109259748A (en) * | 2018-08-17 | 2019-01-25 | 西安电子科技大学 | The system and method for handset processes face video extraction heart rate signal |
| CN111568417A (en) * | 2020-06-01 | 2020-08-25 | 浙江大学 | Non-contact type respiration real-time monitoring method |
| CN112232256A (en) * | 2020-10-26 | 2021-01-15 | 南京读动信息科技有限公司 | Non-contact motion and body measurement data acquisition system |
| CN113052115A (en) * | 2021-04-06 | 2021-06-29 | 合肥工业大学 | Unmanned aerial vehicle airborne vital sign detection method based on video method |
| CN113229790A (en) * | 2021-05-17 | 2021-08-10 | 浙江大学 | Non-contact mental stress assessment system |
| CN113256648A (en) * | 2021-04-30 | 2021-08-13 | 深圳市爱贝宝移动互联科技有限公司 | Self-adaptive multi-scale respiration monitoring method based on camera |
| CN113288058A (en) * | 2021-05-27 | 2021-08-24 | 北京清雷科技有限公司 | Signal processing method and device |
| CN113361526A (en) * | 2021-07-16 | 2021-09-07 | 合肥工业大学 | Non-contact respiration rate monitoring method fusing shoulder and chest area information |
| CN114041767A (en) * | 2021-10-11 | 2022-02-15 | 宁波春建电子科技有限公司 | Heart rate detection method based on depth camera and millimeter wave radar |
| CN115089162A (en) * | 2022-05-30 | 2022-09-23 | 合肥工业大学 | Breathing rate detection method and device based on unmanned aerial vehicle video |
| WO2023068372A1 (en) * | 2021-10-22 | 2023-04-27 | アステラス製薬株式会社 | Thoracic motion measurement device, thoracic motion measurement program, and non-temporary storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100061596A1 (en) * | 2008-09-05 | 2010-03-11 | Varian Medical Systems Technologies, Inc. | Video-Based Breathing Monitoring Without Fiducial Tracking |
| US20130053718A1 (en) * | 2011-08-26 | 2013-02-28 | National Taiwan University | System and method of noncontact and regional respiratory detection using a depth camera |
| CN104173051A (en) * | 2013-05-28 | 2014-12-03 | 天津点康科技有限公司 | Automatic noncontact respiration assessing system and assessing method |
| CN104812300A (en) * | 2012-09-19 | 2015-07-29 | 瑞思迈传感器技术有限公司 | System and method for determining sleep stage |
| CN105007806A (en) * | 2013-03-13 | 2015-10-28 | 皇家飞利浦有限公司 | Apparatus and method for determining vital signs from a subject |
| CN105989357A (en) * | 2016-01-18 | 2016-10-05 | 合肥工业大学 | Human face video processing-based heart rate detection method |
-
2017
- 2017-03-13 CN CN201710145390.5A patent/CN106901741A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100061596A1 (en) * | 2008-09-05 | 2010-03-11 | Varian Medical Systems Technologies, Inc. | Video-Based Breathing Monitoring Without Fiducial Tracking |
| US20130053718A1 (en) * | 2011-08-26 | 2013-02-28 | National Taiwan University | System and method of noncontact and regional respiratory detection using a depth camera |
| CN104812300A (en) * | 2012-09-19 | 2015-07-29 | 瑞思迈传感器技术有限公司 | System and method for determining sleep stage |
| CN105007806A (en) * | 2013-03-13 | 2015-10-28 | 皇家飞利浦有限公司 | Apparatus and method for determining vital signs from a subject |
| CN104173051A (en) * | 2013-05-28 | 2014-12-03 | 天津点康科技有限公司 | Automatic noncontact respiration assessing system and assessing method |
| CN105989357A (en) * | 2016-01-18 | 2016-10-05 | 合肥工业大学 | Human face video processing-based heart rate detection method |
Non-Patent Citations (3)
| Title |
|---|
| DAVIDE ALINOVI ; LUCA CATTANI ; GIANLUIGI FERRARI ; FRANCESCO PI: "Spatio-temporal video processing for respiratory rate estimation", 《2015 IEEE INTERNATIONAL SYMPOSIUM ON MEDICAL MEASUREMENTS AND APPLICATIONS (MEMEA) PROCEEDINGS》 * |
| NEAL WADHWA，MICHAEL RUBINSTEIN,FRE´DO DURAND,WILLIAM T. FREEMAN: "Phase-Based Video Motion Processing", 《ACM TRANSACTIONS ON GRAPHICS》 * |
| 张言飞; 欧阳健飞; 姚丽峰: "基于快速人脸检测的实时呼吸测量系统的设计", 《计算机工程与应用》 * |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107550509A (en) * | 2017-10-16 | 2018-01-09 | 合肥工业大学 | A kind of X-ray machine imaging triggering method based on the detection of contactless breathing pattern |
| CN107967684A (en) * | 2017-12-11 | 2018-04-27 | 深圳大学 | Contactless sleep-respiratory detection method and device |
| CN109063763A (en) * | 2018-07-26 | 2018-12-21 | 合肥工业大学 | Video minor change amplification method based on PCA |
| CN109124613A (en) * | 2018-08-17 | 2019-01-04 | 西安电子科技大学 | Contactless heart rate monitoring unit and method based on interactive voice |
| CN109259748A (en) * | 2018-08-17 | 2019-01-25 | 西安电子科技大学 | The system and method for handset processes face video extraction heart rate signal |
| CN111568417B (en) * | 2020-06-01 | 2021-09-24 | 浙江大学 | Non-contact respiration real-time monitoring method and hardware system adopting same |
| CN111568417A (en) * | 2020-06-01 | 2020-08-25 | 浙江大学 | Non-contact type respiration real-time monitoring method |
| CN112232256A (en) * | 2020-10-26 | 2021-01-15 | 南京读动信息科技有限公司 | Non-contact motion and body measurement data acquisition system |
| CN112232256B (en) * | 2020-10-26 | 2024-02-02 | 南京读动信息科技有限公司 | Contactless sports and body measurement data acquisition system |
| CN113052115A (en) * | 2021-04-06 | 2021-06-29 | 合肥工业大学 | Unmanned aerial vehicle airborne vital sign detection method based on video method |
| CN113256648A (en) * | 2021-04-30 | 2021-08-13 | 深圳市爱贝宝移动互联科技有限公司 | Self-adaptive multi-scale respiration monitoring method based on camera |
| CN113229790A (en) * | 2021-05-17 | 2021-08-10 | 浙江大学 | Non-contact mental stress assessment system |
| CN113288058A (en) * | 2021-05-27 | 2021-08-24 | 北京清雷科技有限公司 | Signal processing method and device |
| CN113361526A (en) * | 2021-07-16 | 2021-09-07 | 合肥工业大学 | Non-contact respiration rate monitoring method fusing shoulder and chest area information |
| CN113361526B (en) * | 2021-07-16 | 2022-09-27 | 合肥工业大学 | Non-contact respiration rate monitoring method fusing shoulder and chest area information |
| CN114041767A (en) * | 2021-10-11 | 2022-02-15 | 宁波春建电子科技有限公司 | Heart rate detection method based on depth camera and millimeter wave radar |
| WO2023068372A1 (en) * | 2021-10-22 | 2023-04-27 | アステラス製薬株式会社 | Thoracic motion measurement device, thoracic motion measurement program, and non-temporary storage medium |
| CN115089162A (en) * | 2022-05-30 | 2022-09-23 | 合肥工业大学 | Breathing rate detection method and device based on unmanned aerial vehicle video |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106901741A (en) | A kind of respiratory rate detection method suitable for environment round the clock | |
| CN103702014B (en) | Non-contact physiological parameter detection method, system and device | |
| CN105190691B (en) | Equipment for obtaining the vital sign of object | |
| CN102499664B (en) | Video-image-based method and system for detecting non-contact vital sign | |
| Iozzia et al. | Relationships between heart-rate variability and pulse-rate variability obtained from video-PPG signal using ZCA | |
| CN106778695B (en) | Multi-person rapid heart rate detection method based on video | |
| CN105007806B (en) | Device and method for determining the vital sign from object | |
| CN113420624B (en) | Non-contact fatigue detection method and system | |
| CN110269600B (en) | Non-contact video heart rate detection method based on multivariate empirical mode decomposition and combined blind source separation | |
| CN105989357A (en) | Human face video processing-based heart rate detection method | |
| CN105962915B (en) | Contactless humanbody respiratory rate and heart rate method for synchronously measuring and system | |
| US20160022204A1 (en) | An apparatus for remote contactless monitoring of sleep apnea | |
| CN103284743B (en) | Determine the method and device of marker location and patient respiratory curve | |
| CN111243739A (en) | Anti-interference physiological parameter telemetering method and system | |
| CN106491117A (en) | A kind of signal processing method and device based on PPG heart rate measurement technology | |
| CN112233813A (en) | Non-contact non-invasive heart rate and respiration measurement method and system based on PPG | |
| CN110276271A (en) | Merge the non-contact heart rate estimation technique of IPPG and depth information anti-noise jamming | |
| CN102309318A (en) | Method for detecting human body physiological parameters on basis of infrared sequence image | |
| CN111839489A (en) | Non-contact physiological and psychological health detection system | |
| CN104173051A (en) | Automatic noncontact respiration assessing system and assessing method | |
| CN108937905B (en) | Non-contact heart rate detection method based on signal fitting | |
| US20200311388A1 (en) | Human body physiological parameter monitoring method based on face recognition for workstation | |
| CN110801212B (en) | BCG signal heart rate extraction method based on neural network | |
| CN110236515A (en) | A kind of contactless heart rate detection method based on near-infrared video | |
| CN110200607A (en) | Method for eliminating body motion influence in vital sign detection based on optical flow method and LMS 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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170630 |