CN104834946B - A kind of contactless sleep monitor method and system - Google Patents
A kind of contactless sleep monitor method and system Download PDFInfo
- Publication number
- CN104834946B CN104834946B CN201510166500.7A CN201510166500A CN104834946B CN 104834946 B CN104834946 B CN 104834946B CN 201510166500 A CN201510166500 A CN 201510166500A CN 104834946 B CN104834946 B CN 104834946B
- Authority
- CN
- China
- Prior art keywords
- sleep
- human body
- information
- pixel
- region
- 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.)
- Active
Links
Landscapes
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The present invention proposes a kind of contactless sleep monitor method and system, and methods described comprises the following steps:The video image in human body head region and chest and abdomen region is gathered using infrared monitoring camera, and utilizes the attitude information of somatosensory device collection human body;Obtain the sleep info of human body;The sleep info of the human body is compared with default standard sleep information, analyzes the health degree of the sleep quality of the human body.The present invention is acquired based on image processing techniques in non-contacting method to sleep quality signal, keeps the natural sleep state of user as far as possible, and the sleep state information of physiological significance is extracted by video image processing technology.And equipment cost is relatively low, it is easy to daily use of households.
Description
Technical field
The present invention relates to sleep monitor technical field, more particularly to a kind of contactless sleep monitor method and system.
Background technology
Sleep quality process is that the brain area neuron presented by the multiple regional interactions of cerebral function with adjusting is slept
Physiological status, be the rest of normal human's function and the important stage recovered, it is in sleep that life, which has 1/3rd time,
Spend, sleep quality is most important to health.Medical research shows in recent years, many major diseases of modern humans,
Such as hypertension, coronary heart disease, arrhythmia cordis, diabetes, the often OSAS with often occurring in sleep
(Sleep Hypopnea Syn-drome, SAH) is relevant.SAH has become a kind of disease for seriously endangering people's life and health.
In addition, significant change can occur for many physiological functions of sleep procedure inside of human body, such as decreased heart rate, drop in blood pressure, new old generation
Thank to slow, respiration rate to reduce.It is a variety of slow due to the change of the continuity and body function regulation system of sleep procedure time
The early signal of venereal disease is prone to catch during this period.Therefore, sleep monitor technology turn into modern medical diagnosis in can not
The content lacked.
The main method of sleep-disorder monitoring at present is to use polysomnogram and the sensitive mattress of fine motion.Polysomnogram leads to
Adhesive electrode is crossed on human body to measure relevant parameter, its monitoring parameters mainly include electroencephalogram, electrocardiogram, electroculogram, under
Neck quite more than 10 physiology such as electromyogram, mouth, nasal respiration air-flow, chest and abdomen respiratory movement, blood oxygen saturation, the sound of snoring, position, tibialis anterior
Signal.Micro-pressure is converted into electric signal by the sensitive mattress of fine motion by pressure sensor, can monitor breathing, and pulse and body are dynamic etc.
Parameter, realize the monitoring of zero load or low-load sleep.But existing sleep-disorder monitoring technology is substantially contact,
They need to carry out in hospital, while need technical professional to operate.This brings very big psychological pressure to patient, also may be used
The measurement result of mistake can be caused, the price of monitoring device is also costly.
Except the method for above-mentioned contact, also a kind of contactless monitoring method based on video and image procossing.This
The equipment that method of the class based on video processing technique is mainly used is thermal infrared imager, but because price is sufficiently expensive, can only
Used in medical research institute, it is difficult to popularize.And due to the limitation of technical conditions, it is difficult to which substantial amounts of data are handled
And excavate useful information.
The content of the invention
The purpose of the present invention is intended at least solve one of described technological deficiency.
Therefore, it is an object of the invention to propose a kind of contactless sleep monitor method and system, pass through video image
Treatment technology extracts the sleep state information of physiological significance.
To achieve these goals, the embodiment of one aspect of the present invention provides a kind of contactless sleep monitor method, bag
Include following steps:
S1, the video image in human body head region and chest and abdomen region is gathered using infrared monitoring camera, and utilizes body
Feel the attitude information of equipment collection human body;
S2, the sleep info of human body is obtained, wherein, the sleep info includes:Information, human body are opened and closed to human eye
Respiration information and sleep procedure in attitude information, comprise the following steps:
The video image of the head zone is analyzed to be monitored and track to head movement, obtain human eye open and
Information, and the video image in analysis chest and abdomen region are closed, the brightness for obtaining the pixel of the chest and abdomen area video image becomes
Change to obtain the respiration information of human body, use default sorting algorithm to classify the attitude information to obtain sleep quality mistake
Attitude information in journey;
S3, the sleep info of the human body is compared with default standard sleep information, analyzes sleeping for the human body
The health degree of dormancy quality.
Contactless sleep monitor method according to embodiments of the present invention, based on image processing techniques in non-contacting method
Sleep quality signal is acquired, the natural sleep state of user is kept as far as possible, is extracted by video image processing technology
There is the sleep state information of physiological significance.Also, present device cost is relatively low, is easy to daily use of households.
Further, information is opened and closed to the human eye that obtains, and comprises the following steps:
Human face region in the area video image of head is initialized;
Stencil matching method and the motion of optical flow tracking method track human faces are respectively adopted to the human face region of initialization;
Stencil matching result and optical flow tracking result are integrated to generate the present frame tracking result of face;
Calculate the coefficient correlation of the present frame tracking result all templates in default ATL, with all coefficients most
It is worth as tracking result and the coefficient correlation of ATL greatly, if the coefficient correlation is more than or equal to first threshold, by institute
State present frame tracking result to be added in ATL, update ATL;Otherwise ATL does not update;
The gray value of all location points in statistical trace results area, the pixel in region is carried out according to Second Threshold
Binaryzation, treat that reconnaissance judges whether in results area position the figure for belonging to eyes eye opening according to more than the Second Threshold
As region, after being screened according to position, pixel still with a grain of salt then judges that eyes to open state, are recognized if no pixel
Closure state is in for eyes.
Further, the respiration information for obtaining human body, comprises the following steps:
The chest and abdomen movable information of the chest and abdomen area video image is amplified, and is converted into the brightness of corresponding pixel points
Change;
Brightness change intense regions are obtained, the brightness change of pixel is converted into breath signal, including:Count the chest
Brightness value is the pixel number sum of maximum or minimum value in abdomen motion intense regions, wherein the change of pixel sum
For respiratory movement when thorax abdomen fluctuations, the undulatory motion change curve obtained according to pixel number is filtered behaviour
Make, generate breath signal;
The breath signal is subjected to Fourier transformation, the frequency on frequency spectrum corresponding to highest amplitude is the breathing of human body
Frequency.
Further, the attitude information obtained during sleep quality, comprises the following steps:
The structure light sent by the somatosensory device obtains human skeleton data;
Mean filter is carried out to the skeleton data to carry out debounce to the skeleton data;
It is trained according to skeleton data after debounce using SVM classifier on training set, the skeleton data is carried out
Classification, determine the sleep attitude information of human body.
The embodiment of another aspect of the present invention provides a kind of contactless sleep monitor system, including:Infrared monitoring camera
Machine, for gathering the video image in human body head region and chest and abdomen region;Somatosensory device, for gathering the attitude information of human body;
Image processing module, for obtaining the sleep info of human body, wherein, the sleep info includes:Letter is opened and closed to human eye
Cease, the attitude information in the respiration information and sleep procedure of human body, including described image processing module analyzes the head zone
Video image to be monitored and track to head movement, obtain human eye opens and closes information, and analysis chest and abdomen area
The video image in domain, the brightness for obtaining the pixel of the chest and abdomen area video image change to obtain the respiration information of human body,
Default sorting algorithm is used to classify the attitude information to obtain the attitude information during sleep quality;Sleep quality
Analysis module, for the sleep info of the human body to be compared with default standard sleep information, analyze the human body
The health degree of sleep quality.
Contactless sleep monitor system according to embodiments of the present invention, based on image processing techniques in non-contacting method
Sleep quality signal is acquired, the natural sleep state of user is kept as far as possible, is extracted by video image processing technology
There is the sleep state information of physiological significance.Also, present device cost is relatively low, is easy to daily use of households.
Further, infrared filter is installed before the infrared monitoring camera.
Further, what described image processing module obtained human eye opens and closes information, including:To head area video figure
Human face region as in is initialized, and stencil matching method and optical flow tracking method is respectively adopted to the human face region of initialization
Track human faces are moved, and stencil matching result and optical flow tracking result are integrated to generate the present frame tracking result of face,
Calculate the coefficient correlation of all templates in the present frame tracking result and default ATL, using maximum in all coefficients as
The coefficient correlation of tracking result and ATL, if the coefficient correlation is more than or equal to first threshold, by the present frame
Tracking result is added in ATL, updates ATL, the gray value of all location points in statistical trace results area, according to the
Pixel in region is carried out binaryzation by two threshold values, treats that reconnaissance is in place in results area institute according to more than the Second Threshold
The image-region for judging whether to belong to eyes eye opening is put, after being screened according to position, pixel still with a grain of salt then judges eyes
To open state, think that eyes are in closure state if no pixel.
Further, described image processing module obtains the respiration information of human body, including:To the chest and abdomen area video image
Chest and abdomen movable information be amplified, and be converted into corresponding pixel points brightness change;Chest and abdomen motion intense regions are obtained, by picture
The brightness change of vegetarian refreshments is converted into breath signal, including:It is most to count brightness value in the pixel brightness change intense regions
The fluctuating of thorax abdomen when the pixel number sum of big value or minimum value, the change of wherein pixel sum turn to respiratory movement
Change, the undulatory motion change curve obtained according to pixel number are filtered operation, generate breath signal;By the breathing
Signal carries out Fourier transformation, and the frequency on frequency spectrum corresponding to highest amplitude is the respiratory rate of human body.
Further, described image processing module obtains the attitude information during sleep quality, comprises the following steps:
The structure light sent by the somatosensory device obtains human skeleton data;
Mean filter is carried out to the skeleton data to carry out debounce to the skeleton data;
It is trained according to skeleton data after debounce using SVM classifier on training set, the skeleton data is carried out
Classification, determine the sleep attitude information of human body.
The additional aspect of the present invention and advantage will be set forth in part in the description, and will partly become from the following description
Obtain substantially, or recognized by the practice of the present invention.
Brief description of the drawings
The above-mentioned and/or additional aspect and advantage of the present invention will become in the description from combination accompanying drawings below to embodiment
Substantially and it is readily appreciated that, wherein:
Fig. 1 is the flow chart according to the contactless sleep monitor method of the embodiment of the present invention;
Fig. 2 is the flow chart according to the sleep info of the acquisition human body of the embodiment of the present invention;
Fig. 3 is the structure chart according to the contactless sleep monitor system of the embodiment of the present invention.
Embodiment
Embodiments of the invention are described below in detail, the example of the embodiment is shown in the drawings, wherein from beginning to end
Same or similar label represents same or similar element or the element with same or like function.Below with reference to attached
The embodiment of figure description is exemplary, it is intended to for explaining the present invention, and is not considered as limiting the invention.
As shown in figure 1, the contactless sleep monitor method of the embodiment of the present invention, comprises the following steps:
Step S1, the video image in human body head region and chest and abdomen region, and profit are gathered using infrared monitoring camera
With the attitude information of somatosensory device collection human body.Wherein, somatosensory device can use Kinect somatosensory device.
Specifically, in this step, active infra-red CCTV camera is used for the video for gathering human body head and thorax abdomen
Image, somatosensory device are used for the overall posture for gathering human body.The wavelength that somatosensory device sends light is 830nm, infrared monitoring camera
The wavelength that machine receives infrared light is generally 780-940nm.Therefore infrared monitoring camera imaging can be sent light by somatosensory device
Interference.In order to ensure the quality of the video data of collection, interference is reduced, it is necessary to be installed before active infra-red CCTV camera
850nm infrared filter.
Step S2, the sleep info of human body is obtained, wherein, sleep info includes:Information, human body are opened and closed to human eye
Respiration information and sleep procedure in attitude information.
As shown in Fig. 2 step S2 comprises the following steps:
Step S21, the video image of head zone is analyzed to be monitored and track to head movement, obtains opening for human eye
Information is closed in open and close.That is, infrared monitoring camera utilizes video interface by image transmitting to image processing module, image procossing mould
Block carries out detect and track to head, so as to which obtain eyes opens and close information.
First, the human face region in the area video image of head is initialized.
Specifically, face position is detected on the first two field picture in head zone video image.It is it is required that tested
Personnel, should be with state front that eyes are opened towards camera lens in the starting stage of video acquisition.Afterwards can be according to tested people
The suitable mode of member is slept.Haar-like features are extracted on the two field picture of video first.Haar-like features are by linear
Feature, edge feature, point feature and diagonal feature composition.In order to accelerate feature extraction speed, calculated using integrogram method
Haar-like features.After extracting feature, face and non-face grader can be distinguished using Adaboost algorithm training,
This grader can be integrated the classification results of one group of Weak Classifier, and strong classifier is concatenated together improving and divided
The accuracy rate of class.The human face region image that the first frame is extracted afterwards is stored in the ATL of face.
Then, stencil matching method is respectively adopted to the human face region of initialization and optical flow tracking method track human faces is transported
It is dynamic.
Specifically, according to the human face region detected, its motion is tracked using two methods of template matches and optical flow tracking
Process is simultaneously integrated result.Be first depending on former frame tracking result determine present frame matching and tracking search model
Enclose.On current frame image, centered on former frame tracking result region, to results area around be extended, gained area
Region for former 9 times of tracing area area is exactly the region of search of present frame.Then.In region of search carry out template matches and
Optical flow tracking.
(1) template matches:
To each template in ATL, the maximum position of coefficient associated therewith is found in region of search using the template
Put, the position is exactly human face region corresponding with search pattern in present frame.Compare face corresponding to all templates in ATL
Region coefficient correlation, the maximum position of coefficient are exactly the result of template matches.
In above process, coefficient correlation of the template in each position is calculated by below equation:
Wherein, R (x, y) is coefficient correlation, represents matching of the target to be tracked in image I coordinate (x, y) place template T
Degree, coefficient correlation is bigger to represent that template and the location drawing picture region similitude are higher.
(2) optical flow tracking:
The selected characteristic point first in the tracking result region of former frame, distance is fixed and is evenly distributed between characteristic point
In whole results area.Then compare the gray scale of present frame and former frame pixel using optical flow method, former frame spy is estimated with this
Sign point is in the position of present frame.The constant characteristic point of strain position, work as according to the direction of motion of remaining characteristic point and apart from determination
The tracking result of previous frame.
Stencil matching result and optical flow tracking result are integrated to generate the present frame tracking result of face.Specifically
Ground, when the area in template matching results region and the coincidence of optical flow tracking results area is more than or equal to selected threshold value, with both
Central point of the intermediate value of tracking result central point transverse and longitudinal coordinate position as the final tracking result region of present frame;Work as overlapping region
When area is less than selected threshold value, favored area is treated as ought closer to former frame final result region using center position in both
The final tracking result of previous frame.
ATL is updated, present frame tracking result is judged, judges whether to be added in ATL.Calculate and work as
The coefficient correlation of previous frame tracking result all templates in default ATL, using maximum in all coefficients as tracking result and
The coefficient correlation of ATL, if coefficient correlation is more than or equal to first threshold, present frame tracking result is added to template
In storehouse, ATL is updated;Otherwise, tracking result is added without in ATL, and ATL does not update.
After the completion of ATL renewal, the status information of eyes is obtained using obtained tracking result.In common infrared prison
Control under video camera, for eyeball due to the reflection of light, what can be showed on image is very bright when human eye is opened.According to this existing
As in tracking result region it may determine that eyes open closed state.The gray scale of all location points in statistical trace results area
Value, the pixel in region is carried out by binaryzation according to Second Threshold, treats reconnaissance in results area according to more than Second Threshold
Position judges whether the image-region for belonging to eyes eye opening, and after being screened according to position, pixel still with a grain of salt is then sentenced
Disconnected eyes think that eyes are in closure state to open state if no pixel.Finally record the eye-shaped of each two field picture
State data.
Step S22, the video image in analysis chest and abdomen region, obtain the brightness change of the pixel of chest and abdomen area video image
To obtain the respiration information of human body.That is, the small movable information of thorax abdomen is amplified, and the brightness for being converted into pixel becomes
Change, choose the violent region of brightness change intensity to obtain the respiration information of human body.
First, the chest and abdomen movable information of chest and abdomen area video image is amplified, and is converted into the bright of corresponding pixel points
Degree change.
Respiration information is extracted according to thorax abdomen undulatory motion.In sleep procedure, thorax abdomen motion when measured breathes
Smaller, frequency is slow, and fluctuating quantity is small, needs the state by chest and abdomen motion to be amplified before obtaining respiration information.Then, root
Respiratory rate curve is extracted according to the motion process of amplification.
Amplify the motion state of thorax abdomen, trickle chest and abdomen motion amplification is changed for obvious pixel brightness.Will
Colored video image has rgb color space to transform to YIQ color spaces.Change between YIQ color spaces and RGB color
It is small to change amount of calculation, and YIQ color spaces are stronger to the continually changing occasion adaptability of intensity of illumination.By under YIQ color spaces
Each channel image carries out laplacian image pyramid decomposition, thus can obtain the multiple dimensioned subgraph for representing piece image
Image set closes.Laplacian pyramid is that LPF is first carried out to original image and down-sampled obtains the approximation of a thick yardstick
Image, that is, obtained low pass approximate image is decomposed, then by this approximate image by interpolation, filtering, then calculate it and artwork
The difference of picture, the band logical component decomposed under the yardstick is just obtained, it is on obtained low pass approximate image that next stage, which decomposes, afterwards
Carry out, iteration completes multi-resolution decomposition.Last in laplacian pyramid layer is the low frequency component of image, and remaining yardstick is then
The high fdrequency component of image.
Image in laplacian pyramid is filtered.By each tomographic image in pyramid and selected wave filter
Convolution is carried out, so as to obtain filtered image.The selection of wave filter is determined by the frequency for the motion for needing to amplify, that is, ensures filter
The frequency that motion is included by frequency range of ripple device.
The brightness amplification of image slices vegetarian refreshments.To every layer in pyramid of image, filtered image pixel value with selecting
Amplification factor α be multiplied, every tomographic image after being just amplified then is added with the image before filtering.Amplification factor α with
The wavelength of moving wave shape is relevant.Its calculation formula is as follows:
Wherein, δ (t) is the displacement of undulation, and λ is wavelength.For amplification factor α when meeting above formula, amplification result is not
Noise is introduced, but magnification level is smaller;When amplification factor is unsatisfactory for above formula, motion amplification degree is larger, but can be knot
Fruit introduces noise.
The layered image of amplification is reassembled into archeus image.By pyramidal multiple dimensioned each tomographic image according to foundation
The reverse order of process is reassembled into the archeus image as before decomposing.Thus obtained image sequence is exactly will motion
It is converted into the result that pixel brightness changes and is amplified.
Secondly, chest and abdomen motion intense regions are obtained, the brightness change of pixel is converted into breath signal, including:Statistics
Brightness value is the pixel number sum of maximum or minimum value in pixel brightness change intense regions, and wherein pixel is total
The fluctuations of thorax abdomen when several changes turns to respiratory movement, the undulatory motion change curve obtained according to pixel number enter
Row filtering operation, generate breath signal.
The brightness of initialized pixel point changes most violent region first.Certain noise information can be introduced after motion amplification,
Disturb the process of signal extraction.Change figure the most violent, it is necessary to select brightness in order to ensure the accuracy of motion state
As region.Entire image is uniformly divided into 9 regions, counts the brightness variance of pixel in each region respectively.Variance compared with
Big region is exactly the more violent region of motion.
After it is determined that brightness changes more violent region, brightness change is converted into breath signal.It is bright in statistical regions
Angle value is the pixel number sum of maximum or minimum value, the thorax abdomen when change of pixel sum is exactly respiratory movement
Fluctuations.Certain noise signal be present, it is necessary to be put down in the undulatory motion change curve obtained according to pixel number
Sliding filtering operation.To the signal value of each frame, each frame signal in certain intervals before and after present frame is counted, takes its average conduct
The signal value of present frame.The length at interval affects the result of filtering, is spaced long respiratory rate and is possible to be filtered, and is spaced
Too short noise remove DeGrain.
Finally, measured's respiratory rate is drawn according to gained respiratory waveform.Breath signal is subjected to Fourier transformation, in frequency
Frequency in spectrum corresponding to highest amplitude is exactly the respiratory rate of measured.
Step S23, default sorting algorithm is used to classify attitude information to obtain the posture during sleep quality
Information.
Structure light is sent by somatosensory device to obtain the data message of human skeleton.In somatosensory device, framework information
It is to be represented with the coordinates of 20 artis.When measured is in the Kinect visuals field, human body attitude is sat with bone node
Target form shows, and when measured moves, corresponding bone node location also changes.Then pass through to framework information
Analysis, it is possible to obtain measured sleep when posture and attitudes vibration information.
First, the structure light sent by somatosensory device obtains human skeleton data.
Specifically, original skeleton data is obtained by somatosensory device.Using the coordinate position of each artis, can obtain
Obtain artis spatial relation.
Then, mean filter is carried out to skeleton data to carry out debounce to skeleton motion track, eliminates skeleton motion track
Shake.
Specifically, in the acquisition stage of skeleton data, it sometimes appear that there is great-jump-forward change in skeleton motion.Cause this
The reason for situation is many, for example testee's arm motion is very fast, Kinect own hardware performance deficiencies etc..Bone closes
The relative position of section changes excessively acutely between frames, considerable influence can be produced to posture analysis later, therefore need
Eliminate the shake of bone node data.For each artis, using the method for mean filter, calculate using present frame in
Bone coordinate data average before and after the heart in certain intervals, the coordinate result using average as present frame.Gained filter result is just
Be it is smooth after skeleton data.
Finally, it is trained according to skeleton data after debounce using SVM classifier on training set, skeleton data is carried out
Classification, determine the sleep attitude information of human body.
Specifically, sleep quality posture carries out preliminary classification first with bone node space relative position.According to arm and
The position relationship of leg node and backbone node, can be to lie low or lie on one's side using the written posture of initial division measured.Lie low
When, arm and leg Node distribution are in backbone node both sides;When lying on one's side, arm and leg Node distribution are in backbone node homonymy.
Simple classification can divide according to the rule of locus.Complex situations are entered by the way of data point is combined with grader
OK.Collect the skeleton node data of several testees and carry out careful classification annotation, in this, as the training number of grader
According to collection.It is trained using SVM classifier on training set.SVM classifier, which is widely used in statistical classification and returned, divides
Analysis, experience error can be minimized simultaneously with maximizing Geometry edge area, solving small sample, the identification of non-linear and high dimensional pattern
In show many distinctive advantages.The skeleton data of current measured is classified using the grader trained, it is determined that
The sleep posture of measured.In addition when skeleton data has motion by a relatively large margin, illustrate measured have in sleep procedure compared with
Significantly body changes, and records now data, finally obtains sleep quality attitude information and motion process.
In one embodiment of the invention, step S21, step S22 and step S23 can be carried out simultaneously.
Step S3, the sleep info of human body is compared with default standard sleep information, analyzes the sleep matter of human body
The health degree of amount.
The measured's sleep info obtained using step S2 is contrasted with the information in standard database, when there is some
Or some signals when having larger difference, there are some problems in the sleep quality of measured;If signal contrast is more or less the same, measured
Sleep quality is more normal.
Contactless sleep monitor method according to embodiments of the present invention, based on image processing techniques in non-contacting method
Sleep quality signal is acquired, the natural sleep state of user is kept as far as possible, is extracted by video image processing technology
There is the sleep state information of physiological significance.Also, present device cost is relatively low, is easy to daily use of households.
As shown in figure 3, the contactless sleep monitor system of the embodiment of the present invention, including:Infrared monitoring camera 1, body
Feel equipment 2, image processing module 3 and Analysis of sleeping quality module 4.
Specifically, infrared monitoring camera 1 is used for the video image for gathering human body head region and chest and abdomen region.Body-sensing is set
Standby 2 attitude information for gathering human body.
In an embodiment of the present invention, infrared filter is installed before infrared monitoring camera 1.Somatosensory device can use
Kinect somatosensory device.
Specifically, the wavelength that somatosensory device 2 sends light is 830nm, and infrared monitoring camera 1 receives the wavelength one of infrared light
As be 780-940nm.Therefore the imaging of infrared monitoring camera 1 can be sent light by somatosensory device and be disturbed.In order to ensure to gather
Video data quality, reduce interference, it is necessary to before active infra-red CCTV camera 1 install 850nm infrared filter.
Image processing module 3 is used for the sleep info for obtaining human body.Wherein, sleep info includes:Human eye opens open and close
Close information, human body respiration information and sleep procedure in attitude information.
Specifically, image processing module 3 analyze head zone video image to be monitored and track to head movement,
Obtain human eye opens and closes information.
What image processing module 3 obtained human eye opens and closes information, including:To the face in the area video image of head
Region is initialized, and stencil matching method is respectively adopted to the human face region of initialization and optical flow tracking method track human faces are transported
It is dynamic, stencil matching result and optical flow tracking result are integrated to generate the present frame tracking result of face, calculate present frame
The coefficient correlation of tracking result all templates in default ATL, tracking result and template are used as using maximum in all coefficients
The coefficient correlation in storehouse, if coefficient correlation is more than or equal to first threshold, present frame tracking result is added in ATL,
ATL is updated, the gray value of all location points in statistical trace results area, according to Second Threshold by the pixel in region
Binaryzation is carried out, treats that reconnaissance judges whether in results area position the figure for belonging to eyes eye opening according to more than Second Threshold
As region, after being screened according to position, pixel still with a grain of salt then judges that eyes to open state, are recognized if no pixel
Closure state is in for eyes.
Image processing module 3 analyzes the video image in chest and abdomen regions, obtain chest and abdomen area video image pixel it is bright
Degree changes to obtain the respiration information of human body.Specifically, image processing module 3 moves letter to the chest and abdomen of chest and abdomen area video image
Breath is amplified, and is converted into the brightness change of corresponding pixel points;Chest and abdomen motion intense regions are obtained, the brightness of pixel is become
Change is converted into breath signal, including:The brightness of statistical pixel point changes the picture that brightness value in strong region is maximum or minimum value
The fluctuations of thorax abdomen when the change of vegetarian refreshments number sum, wherein pixel sum turns to respiratory movement, according to pixel number
The undulatory motion change curve that mesh obtains is filtered smooth operation, generates breath signal;Breath signal is subjected to Fourier's change
Change, the frequency on frequency spectrum corresponding to highest amplitude is the respiratory rate of human body.
During image processing module 3 uses default sorting algorithm to classify attitude information to obtain sleep quality
Attitude information.
The structure light that image processing module 3 is sent by somatosensory device obtains human skeleton data, and skeleton data is entered
Row mean filter is with to skeleton data progress debounce.Carried out according to skeleton data after debounce using SVM classifier on training set
Training, classifies to skeleton data, determines the sleep attitude information of human body.
Analysis of sleeping quality module 4 is used to the sleep info of human body being compared with default standard sleep information, point
Analyse the health degree of the sleep quality of human body.
Contactless sleep monitor system according to embodiments of the present invention, based on image processing techniques in non-contacting method
Sleep quality signal is acquired, the natural sleep state of user is kept as far as possible, is extracted by video image processing technology
There is the sleep state information of physiological significance.Also, present device cost is relatively low, is easy to daily use of households.
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means specific features, structure, material or the spy for combining the embodiment or example description
Point is contained at least one embodiment or example of the present invention.In this manual, to the schematic representation of above-mentioned term not
Necessarily refer to identical embodiment or example.Moreover, specific features, structure, material or the feature of description can be any
One or more embodiments or example in combine in an appropriate manner.
Although embodiments of the invention have been shown and described above, it is to be understood that above-described embodiment is example
Property, it is impossible to limitation of the present invention is interpreted as, one of ordinary skill in the art is not departing from the principle and objective of the present invention
In the case of above-described embodiment can be changed within the scope of the invention, change, replace and modification.The scope of the present invention
Extremely equally limited by appended claims.
Claims (9)
- A kind of 1. contactless sleep monitor method, it is characterised in that comprise the following steps:S1, the video image in human body head region and chest and abdomen region is gathered using infrared monitoring camera, and is set using body-sensing The attitude information of standby collection human body;S2, the sleep info of human body is obtained, wherein, the sleep info includes:Human eye open and close information, human body is exhaled The attitude information in information and sleep procedure is inhaled, is comprised the following steps:The video image of the head zone is analyzed to be monitored and track to head movement, obtains opening and closing for human eye Information, and the video image in analysis chest and abdomen region, using motion amplification algorithm to obtain the respiration information of human body, using default Sorting algorithm is classified to the attitude information of the human body to obtain the attitude information during sleep quality;S3, the sleep info of the human body is compared with default standard sleep information, analyzes the sleep matter of the human body The health degree of amount.
- 2. contactless sleep monitor method as claimed in claim 1, it is characterised in that the acquisition human eye opens open and close Information is closed, is comprised the following steps:Human face region in the video image of head zone is initialized;Stencil matching method and the motion of optical flow tracking method track human faces are respectively adopted to the human face region of initialization;Stencil matching result and optical flow tracking result are integrated to generate the present frame tracking result of face;The coefficient correlation of the present frame tracking result and all templates in default ATL is calculated, with maximum in all coefficients As tracking result and the coefficient correlation of ATL, if the tracking result and the coefficient correlation of ATL are more than or equal to the One threshold value, then the present frame tracking result is added in ATL, updates ATL;Otherwise ATL does not update;The gray value of all pixels point in statistical trace results area, according to Second Threshold by the pixel in tracking result region Binaryzation is carried out, judges whether that belonging to eyes opens eyes in results area position according to the pixel more than the Second Threshold Image-region, after being screened according to position, pixel still with a grain of salt then judges eyes to open state, if without pixel Then think that eyes are in closure state.
- 3. contactless sleep monitor method as claimed in claim 1, it is characterised in that the breathing letter for obtaining human body Breath, comprises the following steps:The chest and abdomen movable information of the video image in the chest and abdomen region is amplified, and the brightness for being converted into corresponding pixel points becomes Change;Brightness change intense regions are obtained, the brightness change of pixel is converted into breath signal, including:The brightness is counted to become Change the pixel sum that brightness value in intense regions is maximum or minimum value, the change of wherein pixel sum turns to breathing fortune The fluctuations of thorax abdomen when dynamic, the undulatory motion change curve obtained according to pixel sum are filtered operation, generate Breath signal;The breath signal is subjected to Fourier transformation, the frequency on frequency spectrum corresponding to highest amplitude is the breathing frequency of human body Rate.
- 4. contactless sleep monitor method as claimed in claim 1, it is characterised in that during the acquisition sleep quality Attitude information, comprise the following steps:The structure light sent by the somatosensory device obtains human skeleton data;Mean filter is carried out to the skeleton data to carry out debounce to the skeleton data;It is trained according to skeleton data after debounce using SVM classifier on training set, skeleton data after the debounce is entered Row classification, determines the attitude information during sleep quality.
- A kind of 5. contactless sleep monitor system, it is characterised in that including:Infrared monitoring camera, for gathering the video image in human body head region and chest and abdomen region;Somatosensory device, for gathering the attitude information of human body;Image processing module, for obtaining the sleep info of human body, wherein, the sleep info includes:Human eye opens open and close Close information, human body respiration information and sleep procedure in attitude information, including described image processing module analyzes the head For the video image in region to be monitored and track to head movement, obtain human eye opens and closes information, and analysis chest The video image in abdomen region, the brightness for obtaining the pixel of the video image in the chest and abdomen region change to obtain the breathing of human body Information, use default sorting algorithm to classify the attitude information of the human body and believed with obtaining the posture during sleep quality Breath;Analysis of sleeping quality module, for the sleep info of the human body to be compared with default standard sleep information, point Analyse the health degree of the sleep quality of the human body.
- 6. contactless sleep monitor system as claimed in claim 5, it is characterised in that pacify before the infrared monitoring camera Equipped with infrared filter.
- 7. contactless sleep monitor system as claimed in claim 5, it is characterised in that described image processing module obtains people Information is opened and closed to eye, including:Human face region in the video image of head zone is initialized, to initialization Stencil matching method and the motion of optical flow tracking method track human faces is respectively adopted in human face region, by stencil matching result and light stream with Track result is integrated to generate the present frame tracking result of face, is calculated in the present frame tracking result and default ATL The coefficient correlation of all templates, using maximum in all coefficients as tracking result and the coefficient correlation of ATL, if described Tracking result and the coefficient correlation of ATL are more than or equal to first threshold, then the present frame tracking result are added into template In storehouse, ATL is updated, the gray value of all pixels point in statistical trace results area, according to Second Threshold by tracking result area Pixel in domain carries out binaryzation, is judged whether according to the pixel more than the Second Threshold in results area position Belong to the image-region of eyes eye opening, after being screened according to position, pixel still with a grain of salt then judges eyes to open state, Think that eyes are in closure state if no pixel.
- 8. contactless sleep monitor system as claimed in claim 5, it is characterised in that described image processing module obtains people The respiration information of body, including:The chest and abdomen movable information of the video image in the chest and abdomen region is amplified, and is converted into correspondingly The brightness change of pixel;Brightness change intense regions are obtained, the brightness change of pixel is converted into breath signal, including: Count the pixel sum that brightness value in the brightness change intense regions is maximum or minimum value, wherein pixel sum Thorax abdomen of change when turning to respiratory movement fluctuations, the undulatory motion change curve obtained according to pixel sum carries out Filtering operation, generate breath signal;The breath signal is subjected to Fourier transformation, the frequency on frequency spectrum corresponding to highest amplitude Rate is the respiratory rate of human body.
- 9. contactless sleep monitor system as claimed in claim 5, it is characterised in that described image processing module obtains people Attitude information during somatic sleep, comprises the following steps:The structure light sent by the somatosensory device obtains human skeleton data;Mean filter is carried out to the skeleton data to carry out debounce to the skeleton data;It is trained according to skeleton data after debounce using SVM classifier on training set, skeleton data after the debounce is entered Row classification, determines the attitude information during sleep quality.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510166500.7A CN104834946B (en) | 2015-04-09 | 2015-04-09 | A kind of contactless sleep monitor method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510166500.7A CN104834946B (en) | 2015-04-09 | 2015-04-09 | A kind of contactless sleep monitor method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104834946A CN104834946A (en) | 2015-08-12 |
CN104834946B true CN104834946B (en) | 2018-02-09 |
Family
ID=53812822
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510166500.7A Active CN104834946B (en) | 2015-04-09 | 2015-04-09 | A kind of contactless sleep monitor method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104834946B (en) |
Families Citing this family (34)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11051765B2 (en) | 2015-12-31 | 2021-07-06 | Shanghai Oxi Technology Co., Ltd | Health status detecting system and method for detecting health status |
WO2017141976A1 (en) * | 2016-02-15 | 2017-08-24 | ヘルスセンシング株式会社 | Device and method for measuring sleep state, phase coherence calculation device, body vibration signal measurement device, stress level mesaurement device, sleep state measurement device, and cardiac waveform extraction method |
CN105869144B (en) * | 2016-03-21 | 2018-10-19 | 常州大学 | A kind of contactless monitoring of respiration method based on depth image data |
CN105760861B (en) * | 2016-03-29 | 2020-05-12 | 华东师范大学 | Epileptic seizure monitoring method and system based on depth data |
CN106388780A (en) * | 2016-09-21 | 2017-02-15 | 广州视源电子科技股份有限公司 | A sleep state detection method and system based on fusion of two classifiers and detectors |
CN106778617B (en) * | 2016-12-17 | 2020-09-11 | 天津工业大学 | Sleep monitoring method and system based on image acquisition technology |
CN106647544A (en) * | 2017-02-23 | 2017-05-10 | 王森 | Intelligent control method and device for electric appliances |
CN107028427B (en) * | 2017-06-08 | 2020-02-21 | 翼猫科技发展(上海)有限公司 | Novel sensing method of sleep system capable of automatically and intelligently adjusting |
CN107316322A (en) | 2017-06-27 | 2017-11-03 | 上海智臻智能网络科技股份有限公司 | Video tracing method and device and object identifying method and device |
CN107174239A (en) * | 2017-07-05 | 2017-09-19 | 李震中 | A kind of sleep monitor |
CN107967684A (en) * | 2017-12-11 | 2018-04-27 | 深圳大学 | Contactless sleep-respiratory detection method and device |
US11664104B2 (en) | 2017-12-21 | 2023-05-30 | Visionhealth Gmbh | Inhaler training system and method |
TWI667635B (en) * | 2018-01-09 | 2019-08-01 | 國立交通大學 | Monitoring system and monitoring method for infant |
CN108433727A (en) * | 2018-03-15 | 2018-08-24 | 广东工业大学 | A kind of method and device of monitoring baby breathing |
CN108272448B (en) * | 2018-03-29 | 2020-11-27 | 合肥工业大学 | Non-contact type infant physiological parameter day and night monitoring method |
CN108629299B (en) * | 2018-04-24 | 2021-11-16 | 武汉幻视智能科技有限公司 | Long-time multi-target tracking method and system combining face matching |
CN108523901A (en) * | 2018-05-14 | 2018-09-14 | 北京工业大学 | A kind of sleep quality monitoring method based on smart mobile phone |
CN108985180A (en) * | 2018-06-22 | 2018-12-11 | 张小勇 | Data processing method and system |
CN109303548A (en) * | 2018-07-20 | 2019-02-05 | 渝新智能科技(上海)有限公司 | A kind of children's sleeping position monitoring method, device and equipment |
EP3628213A1 (en) * | 2018-09-25 | 2020-04-01 | Koninklijke Philips N.V. | Deriving information about a person's sleep and wake states from a sequence of video frames |
CN109431681B (en) * | 2018-09-25 | 2023-12-19 | 吉林大学 | Intelligent eye patch for detecting sleep quality and detection method thereof |
CN110942824A (en) * | 2019-11-21 | 2020-03-31 | 北京健康有益科技有限公司 | Sleep quality evaluation system based on artificial intelligence |
US20230022206A1 (en) * | 2019-12-17 | 2023-01-26 | Shenzhen Institutes Of Advanced Technology | Infrared image sequence-based sleep quality evaluation system and method |
CN111063438B (en) * | 2019-12-17 | 2023-04-18 | 深圳先进技术研究院 | Sleep quality evaluation system and method based on infrared image sequence |
CN112151171A (en) * | 2019-12-23 | 2020-12-29 | 淮安信息职业技术学院 | Physique test method based on identity recognition |
CN111248868A (en) * | 2020-02-20 | 2020-06-09 | 长沙湖湘医疗器械有限公司 | Quick eye movement sleep analysis method, system and equipment |
CN111387957B (en) * | 2020-03-13 | 2023-03-24 | 智方达(天津)科技有限公司 | Non-contact type body temperature and respiration rate combined detection method |
CN112353168B (en) * | 2020-12-09 | 2022-08-23 | 江苏卧尔康家居用品有限公司 | Intelligent voice-control electric mattress |
CN112580522A (en) * | 2020-12-22 | 2021-03-30 | 北京每日优鲜电子商务有限公司 | Method, device and equipment for detecting sleeper and storage medium |
CN112842266B (en) * | 2020-12-31 | 2024-05-14 | 湖南正申科技有限公司 | Sleep stage identification method based on human body monitoring sleep data |
CN112806966B (en) * | 2021-02-03 | 2022-07-26 | 华南理工大学 | Non-interference type early warning system for sleep apnea |
CN113516095A (en) * | 2021-07-28 | 2021-10-19 | 宁波星巡智能科技有限公司 | Infant sleep monitoring method, device, equipment and medium |
CN113693588A (en) * | 2021-08-27 | 2021-11-26 | 骆飞 | Sleep management system |
CN114469005B (en) * | 2022-02-17 | 2023-10-24 | 珠海格力电器股份有限公司 | Sleep state monitoring method and device and computer readable storage medium |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN202942113U (en) * | 2012-11-29 | 2013-05-22 | 中国人民解放军第四军医大学 | Sleep respiratory function monitoring system based on infrared radiation detection |
CN103366505A (en) * | 2013-06-26 | 2013-10-23 | 安科智慧城市技术(中国)有限公司 | Sleeping posture identification method and device |
CN104083460A (en) * | 2014-06-10 | 2014-10-08 | 王春芳 | Traditional Chinese medicine preparation for treating rheumatoid arthritis |
-
2015
- 2015-04-09 CN CN201510166500.7A patent/CN104834946B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN202942113U (en) * | 2012-11-29 | 2013-05-22 | 中国人民解放军第四军医大学 | Sleep respiratory function monitoring system based on infrared radiation detection |
CN103366505A (en) * | 2013-06-26 | 2013-10-23 | 安科智慧城市技术(中国)有限公司 | Sleeping posture identification method and device |
CN104083460A (en) * | 2014-06-10 | 2014-10-08 | 王春芳 | Traditional Chinese medicine preparation for treating rheumatoid arthritis |
Non-Patent Citations (4)
Title |
---|
automated detection of newborn sleep apnea using video monitoring system;shashank sharma et al;《2015 eighth international conference on advance in pattern recognition(ICAPR)》;20150104;第1-6页 * |
BREATH AND POSITION MONITORING;Meng-Chieh Yu et al;《In Proceedings of the International Conference on Health Informatics (HEALTHINF-2012)》;20121231;第12-22页 * |
graph-based depth video denoising and event detection for sleep monitoring;cheng yang et al;《2014 IEEE 16th international workshop on multimedia signal processing(MMSP)》;20140922;第1-6页 * |
红外热像视频的细微变化放大;付传卿 等;《中国图象图形学报》;20141130;第19卷(第11期);第1577-1583页 * |
Also Published As
Publication number | Publication date |
---|---|
CN104834946A (en) | 2015-08-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104834946B (en) | A kind of contactless sleep monitor method and system | |
US11350850B2 (en) | Systems and methods for video-based monitoring of vital signs | |
Sathyanarayana et al. | Vision-based patient monitoring: a comprehensive review of algorithms and technologies | |
Jindal et al. | An adaptive deep learning approach for PPG-based identification | |
US7640055B2 (en) | Self-adaptive system for the analysis of biomedical signals of a patient | |
Wang et al. | Unconstrained video monitoring of breathing behavior and application to diagnosis of sleep apnea | |
CN104545888B (en) | A kind of sleep apnea acquisition analysis system based on dynamic electrocardiogram and respiratory wave acquisition | |
US20140275829A1 (en) | Sleep stage annotation device | |
Cheng et al. | Heterogeneous recurrence analysis of heartbeat dynamics for the identification of sleep apnea events | |
EP3060118A1 (en) | Motion monitor | |
WO2013165550A9 (en) | Methods and systems for measuring dynamic changes in the physiological parameters of a subject | |
Awais et al. | A hybrid DCNN-SVM model for classifying neonatal sleep and wake states based on facial expressions in video | |
CN104545818A (en) | Sleep apnea syndrome detection method based on pulse and blood oxygen signals | |
CN112806966B (en) | Non-interference type early warning system for sleep apnea | |
CN113257440A (en) | ICU intelligent nursing system based on patient video identification | |
CN113633260A (en) | Multi-sleep monitoring method, monitor, computer device and readable storage medium | |
Abbod et al. | Survey on the use of smart and adaptive engineering systems in medicine | |
Gerla et al. | Multivariate analysis of full-term neonatal polysomnographic data | |
Kau et al. | Pressure-sensor-based sleep status and quality evaluation system | |
KR101757431B1 (en) | Method for extracting heart information of facial micro-movement and system adopting the method | |
CN114998229B (en) | Non-contact sleep monitoring method based on deep learning and multi-parameter fusion | |
Mozafari et al. | Comparison of blind source separation techniques for respiration rate estimation from depth video | |
Su et al. | Sleep comfort analysis using a part-based mixture model with nighttime infrared video | |
CN210494064U (en) | Dynamic electrocardio, respiration and motion monitoring equipment | |
Shinde et al. | Facial features based prediction of Parkinson's disease |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
EXSB | Decision made by sipo to initiate substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |