Non-contact blood oxygen saturation measurement device and non-contact blood oxygen saturation measurement and correction method

Abstract

A non-contact blood oxygen saturation measurement device and a non-contact blood oxygen saturation measurement and correction method. The measurement device includes a camera and a processor. The camera includes a first light sensing module and a second light sensing module. The first light sensing module is used to capture ambient light parameters, and the second light sensing module is used to capture target images. The processor is used to perform the following steps: obtaining a color corrected image according to the target image and the ambient light parameters; detecting a region of interest in the color corrected image; obtaining a first color light change signal, a second color light change signal and a third color light change signal of a plurality of different color lights according to the region of interest; obtaining a light ratio parameter according to the first color light change signal, the second color light change signal and the third color light change signal; and obtaining a blood oxygen saturation according to the light ratio parameter.

Description

Technical Field

[0001] A non-contact blood oxygen saturation measuring device and a non-contact blood oxygen saturation measuring and correction method are disclosed, particularly a non-contact blood oxygen saturation measuring and correction method that can be adapted to different environments for estimating the current blood oxygen saturation of a target person in a non-contact manner using a target image. Background Technology

[0002] In recent years, the demand for health monitoring of blood oxygen saturation has received increasing attention, and it has become an important physiological indicator for daily self-health assessment by the general public. Blood oxygen saturation is defined as the proportion of oxyhemoglobin to total hemoglobin in the human body. Currently, most home measurement products on the market use photoplethysmography (PPG) for signal acquisition and parameter calculation, such as fingertip pulse oximeters or sports watches. However, the light signal ratio parameter on which this method is based is easily interfered with by changes in external ambient light, thus affecting the accuracy of the measurement. Therefore, existing blood oxygen measurement devices are almost all contact-type designs, isolating ambient light interference through close contact. Although some academic papers and previous patents have disclosed the concept and architecture of non-contact blood oxygen saturation measurement, these published technical documents have not disclosed effective means to suppress and compensate for interference from changes in ambient light, resulting in non-contact measurement being unable to operate stably and accurately under different home lighting conditions, thus limiting its practicality. Summary of the Invention

[0003] In view of this, in some embodiments, a non-contact blood oxygen saturation measurement device is provided, which includes a camera device and a processor. The camera device includes a first photosensitive module and a second photosensitive module. The first photosensitive module is used to capture ambient light parameters, and the second photosensitive module is used to capture a target image. The processor is used to perform the following steps: obtaining a color-corrected image based on the target image and the ambient light parameters; detecting a region of interest in the color-corrected image; obtaining a first color light change signal, a second color light change signal, and a third color light change signal of multiple different colors of light based on the region of interest; obtaining a light ratio parameter based on the first color light change signal, the second color light change signal, and the third color light change signal; and obtaining blood oxygen saturation based on the light ratio parameter.

[0004] In some embodiments, the processor executes a color correction model, with ambient light parameters input into the color correction model to obtain predicted correction coefficients, and the processor obtains a color-corrected image based on the target image and the predicted correction coefficients.

[0005] In some embodiments, the color correction model is trained with laboratory light sources and multiple control ambient light sources to output predicted correction coefficients, which are either color matrix gain coefficients or three primary color gain coefficients.

[0006] In some embodiments, the color correction model is a convolutional neural network (CNN) model.

[0007] In some embodiments, the first color light change signal includes a first color light DC component and a first color light AC component; the second color light change signal includes a second color light DC component and a second color light AC component; and the third color light change signal includes a third color light DC component and a third color light AC component.

[0008] In some embodiments, the processor is used to obtain the light ratio parameter according to the light ratio estimation formula, which is:

[0009]

[0010] Wherein, RoR is the light ratio parameter; AC1 is the first color light AC component; DC1 is the first color light DC component; AC2 is the second color light AC component; DC2 is the second color light DC component; AC3 is the third color light AC component; DC3 is the third color light DC component; α, β, m, and n are constants associated with the second photosensitive module.

[0011] In some embodiments, the region of interest includes a red light signal, a green light signal, and a blue light signal; the processor obtains a first color change signal, a second color change signal, and a third color change signal corresponding to these colors based on the red light signal, the green light signal, and the blue light signal.

[0012] In some embodiments, the first photosensitive module has a plurality of filter arrays, which respectively include a first color light filter, a second color light filter and a third color light filter; the processor obtains a red color light signal, a green color light signal and a blue color light signal respectively based on the first color light filter, the second color light filter and the third color light filter.

[0013] In some embodiments, the first photosensitive module is a multispectral sensor, and the second photosensitive module is an RGB photosensitive module.

[0014] In some embodiments, a non-contact method for measuring and correcting blood oxygen saturation is also provided, which can effectively overcome the problem of light signal interference caused by changes in ambient light sources. The method includes: acquiring a target image and ambient light parameters; obtaining a color-corrected image based on the target image and ambient light parameters; detecting a region of interest in the color-corrected image; obtaining a first color light change signal, a second color light change signal, and a third color light change signal based on the region of interest; obtaining a light ratio parameter based on the first color light change signal, the second color light change signal, and the third color light change signal; and obtaining blood oxygen saturation based on the light ratio parameter.

[0015] In some embodiments, the step of capturing the target image and ambient light parameters further includes: executing a color correction model, inputting the ambient light parameters into the color correction model to obtain predicted correction coefficients; and obtaining a color-corrected image based on the target image and the predicted correction coefficients.

[0016] In summary, in some embodiments of the non-contact blood oxygen saturation measurement device, ambient light parameters are captured by a first photosensitive module and a target image is captured by a second photosensitive module to estimate blood oxygen saturation. Specifically, the processor can obtain a prediction correction coefficient based on the ambient light parameters and use the prediction correction coefficient to perform a color correction mechanism on the target image, so that the color of the image captured by the camera device under the current ambient light source is consistent with the color of the laboratory ambient light source. Thus, when the processor estimates blood oxygen saturation using the color-corrected image, it can avoid calculation errors in the estimated blood oxygen saturation caused by differences between the ambient light source and the laboratory ambient light source. In other words, the non-contact blood oxygen saturation measurement device can, under varying and arbitrary everyday ambient light conditions, eliminate the influence of environmental noise to obtain accurate light ratio parameters through the signals (such as ambient light parameters and target image) captured by the first and second photosensitive modules and the processor's calculations, thereby achieving real-time non-contact blood oxygen saturation estimation unaffected by ambient light. Accordingly, the non-contact blood oxygen saturation measuring device and its measurement and calibration method in some embodiments can improve the limitations of traditional contact measurement and solve the technical bottleneck that existing non-contact technologies cannot resist ambient light interference.

[0017] The following detailed description of the features and advantages of the present invention in the embodiments is sufficient to enable any person skilled in the art to understand the technical content of the present invention and implement it accordingly. Based on the content disclosed in this specification, the claims and the drawings, any person skilled in the art can easily understand the related objects and advantages of the present invention. Attached Figure Description

[0018] Figure 1 This is a block diagram of a non-contact blood oxygen saturation measuring device in some embodiments of the present invention.

[0019] Figure 2 The flowchart (I) shows a non-contact blood oxygen saturation measurement and correction method in some embodiments of the present invention.

[0020] Figure 3 The flowchart (II) shows a non-contact blood oxygen saturation measurement and correction method in some embodiments of the present invention.

[0021] Figure 4 This is a schematic diagram of blood oxygen measurement using a non-contact blood oxygen saturation measuring device in some embodiments of the present invention, showing the camera device capturing ambient light parameters and target images.

[0022] Figure 5 This is a schematic diagram of blood oxygen measurement using a non-contact blood oxygen saturation measuring device in some embodiments of the present invention, shown in the target image with the region of interest marked.

[0023] Figure 6 This is a schematic diagram of the implementation of the non-contact blood oxygen saturation measurement device in some embodiments of the present invention, showing that the second photosensitive module has multiple filter arrays.

[0024] In the attached figures, the following labels are used:

[0025] 100: Non-contact blood oxygen saturation measurement device

[0026] 102: Camera device

[0027] 104: Processor

[0028] 106: First photosensitive module

[0029] 108: Second photosensitive module

[0030] 110: Storage Module

[0031] 112: Filter array

[0032] 114: First color filter

[0033] 116: Second color light filter

[0034] 118: Third color light filter

[0035] A1: Region of Interest

[0036] S: Non-contact method for measuring and correcting blood oxygen saturation.

[0037] S0, S1, S11, S12, S2, S3, S4, S5: Steps

[0038] TI: Target Image

[0039] Lr: Red light

[0040] Lg: Green light

[0041] Lb: Blue light Detailed Implementation

[0042] In some embodiments, please refer to the following: Figure 1 , Figure 2 , Figure 3 , Figure 4 and Figure 5 The non-contact blood oxygen saturation measuring device 100 includes a camera device 102 and a processor 104.

[0043] The camera device 102 may be, for example, but not limited to, a device with camera functionality. Examples include a camera, a wearable device with a camera lens, a mobile communication device, or a computer. The camera device 102 includes a first photosensitive module 106 and a second photosensitive module 108. The first photosensitive module 106 is used to capture an ambient light parameter. The ambient light parameter may be, for example, at least one or a combination of two or more parameters such as light intensity, spectral distribution, and color temperature. The second photosensitive module 108 is used to capture a target image TI. The target image TI is an image composed of RGB spectra and captured facing the human body. The second photosensitive module 108 can filter out images of the person other than those with RGB spectra to obtain the target image TI. When the camera device 102 performs a shooting action, it can actuate the first photosensitive module 106 and the second photosensitive module 108.

[0044] In some embodiments, the first photosensitive module 106 may be, for example, but not limited to, a multispectral sensor. A multispectral sensor can refer to a sensor whose spectral waveform peak does not overlap with the peak of the three primary colors (RGB) waveform. The multispectral sensor may be, for example, but not limited to, a spectral sensing element (with at least three optical spectral bands) based on the CIE 1931 XYZ tristimulus values ​​standard or AS7341 type, to capture multispectral color parameters, and its spectral range may cover visible light. The second photosensitive module 108 may be, for example, an RGB sensor. The first photosensitive module 106 and the second photosensitive module 108 can convert the captured spectral signals into electronic signals (defining the color distribution of the signal in a standard color space) for transmission to the processor 104 for analysis and processing.

[0045] In some embodiments, the multispectral image sensor includes a color filter array (CFA) and a photosensitive element (such as a charge-coupled device (CCD) or complementary metal-oxide-semiconductor (CMOS)). The color filter array, located above the photosensitive area of ​​the photosensitive element, consists of multiple arrayed pixels, each pixel corresponding to a color filter or color filter coating. Thus, the multispectral image sensor can have multi-dimensional resolution of the spectrum (e.g., 4×4 pixels constitute a 16-dimensional resolution). The multispectral image sensor can convert the acquired color parameters (such as light intensity, spectral distribution, color temperature, tristimulus curves, spectral response function (SRF), and / or reflectance) into digital signals for storage and application by the processor 104.

[0046] Processor 104 is used to acquire target image TI and ambient light parameters to execute a non-contact blood oxygen saturation measurement and correction method S. Processor 104 can correct the target image TI based on the ambient light parameters and estimate the target's blood oxygen saturation using the corrected target image TI. Processor 104 is communicatively connected to a first photosensitive module 106 and a second photosensitive module 108, respectively, to receive signals (which may refer to ambient light parameters) acquired from the first photosensitive module 106 and the target image TI received from the second photosensitive module 108 in real time, and then perform color correction and estimate blood oxygen saturation using the target image TI. Processor 104 and the second photosensitive module 108 can be disposed in the camera device 102, or processor 104 can be a separate and independent device from the camera device 102. Processor 104 can be, for example, but not limited to, a graphics processing unit (GPU), a central processing unit (CPU), or a tensor processing unit (TPU).

[0047] Non-contact blood oxygen saturation measurement and correction method S includes:

[0048] Extract the target image TI and ambient light parameters (step S0);

[0049] A color-corrected image is obtained based on the target image TI and ambient light parameters (step S1);

[0050] A region of interest A1 is detected in the color-corrected image (step S2).

[0051] Based on the region of interest A1, obtain a first color light change signal, a second color light change signal, and a third color light change signal of multiple different colors of light (step S3).

[0052] A light ratio parameter is obtained based on the first color light change signal, the second color light change signal, and the third color light change signal (step S4); and

[0053] Blood oxygen saturation is obtained based on the light ratio parameter (step S5).

[0054] In step S1, during the factory calibration phase, the camera device 102 performs image color calibration under a laboratory ambient light source to adjust the image color to be close to the standard color. Under different ambient light sources, the light color intensity (e.g., red, green, or blue light) sensed by the second photosensitive module 108 will deviate, causing the captured image color to differ from the standard color (this could refer to color cast). It should be noted that the parameters used by the processor 104 to estimate blood oxygen saturation (which could refer to the light ratio parameter or a constant in the light ratio estimation formula described later) are calculated under laboratory ambient light sources. If the image color is inconsistent with the standard color, the error between the blood oxygen saturation estimated by the processor 104 and the actual blood oxygen saturation will increase. Therefore, the processor 104 can correct the image color of the target image TI captured under different ambient light sources to be close to or consistent with the standard color (the laboratory ambient light source is the standard color) based on the ambient light parameters, so that the processor 104 can use the light ratio parameter obtained from the laboratory ambient light source to be applicable to different ambient light sources, thereby overcoming the problem of light signal interference caused by changes in ambient light sources.

[0055] In some embodiments, such as Figure 3 As shown. Step S1 further includes: processor 104 executing a color correction model, inputting ambient light parameters into the color correction model to obtain predicted correction coefficients (step S11); and processor 104 obtaining a color-corrected image based on the target image TI and the predicted correction coefficients (step S12). When the camera device 102 performs a shooting action, processor 104 can input the ambient light parameters captured by the first photosensitive module 106 into the color correction model to obtain predicted correction coefficients. Then, processor 104 uses these predicted correction coefficients to perform a color correction mechanism on the target image TI to generate a color-corrected image. Here, the color of this color-corrected image will be corrected to be close to the color values ​​of the laboratory ambient light source, which can overcome the light signal interference problem caused by changes in the ambient light source, thereby reducing the error in subsequent blood oxygen saturation estimation. It should be noted that the first photosensitive module 106 can perform a preprocessing (or preprocessing through processor 104) on the captured image signal to obtain ambient light parameters. The ambient light parameters include color parameters of three or more spectra.

[0056] The color correction mechanism can be white balance correction and / or color correction matrix correction. White balance correction can be, for example, but not limited to, automatic white balance, which can be based on white balance algorithms such as gray world assumption, white point statistics, or color temperature estimation. The processor 104 can perform white balance correction using the target image TI (or an image captured in the ambient light field) sensed by the second photosensitive module 108 and the prediction correction coefficients. The processor 104 can also perform color matrix correction using the target image TI and the prediction correction coefficients, correcting the color (which can refer to RGB values) of the target image TI to the color sensed under laboratory ambient light (assuming the laboratory ambient light is a standard color). In other words, the processor 104 can adjust the color space of the target image TI to the color space corresponding to the prediction correction coefficients. The processor 104 can preset white balance correction and / or color matrix correction as the default color correction mechanism.

[0057] In some embodiments, the color correction model can be trained using a laboratory light source and multiple control ambient light sources to output predicted correction coefficients. The predicted correction coefficients are either a color matrix gain coefficient or a primary color gain coefficient, enabling the processor 104 to perform color correction mechanisms such as white balance correction and / or color matrix correction. The laboratory light source can refer to ambient light parameters (such as color values ​​in the spectral distribution) sensed under laboratory ambient light sources (such as a D65 light source). The difference between the color matrix gain coefficient and the primary color gain coefficient lies in the fact that the color correction model produces different gain coefficient results during the training phase based on the set color correction mechanism. The control ambient light sources can refer to ambient light sources photographed in different environments; these control ambient light sources can be converted into ambient light parameters to establish a training set for the color correction model.

[0058] In some embodiments, the color correction model may be, for example, a convolutional neural network (CNN) model. The model training process is illustrated using a CNN as an example. Ambient light parameters from multiple control ambient light sources can be input into the untrained color correction model, with the ambient light parameters from the laboratory light source used as labels. During the training phase, the color correction model can predict the ambient light parameter shift characteristics of the current ambient light source and compare them with the labels to repeatedly adjust the model weights. After the color correction model is trained, the trained model can output predicted correction coefficients (color matrix gain coefficients or primary color gain coefficients) based on the input ambient light parameters. It should be noted that, depending on the color correction mechanism implemented, the color correction model can decide during training to output either color matrix gain coefficients or primary color gain coefficients. The red light gain coefficient adjusts the red light of the target image TI, and the blue light gain coefficient adjusts the blue light of the target image TI. Thus, by adjusting the red light gain coefficient and blue light gain coefficient respectively, the target image TI can be adjusted to approximate the color of the laboratory light source. In some embodiments, the termination condition of the color correction model may be the convergence of the corresponding target loss function or the reaching of a default number of iterations. The target loss function and its convergence condition, or the number of iterations, can be specifically set as needed.

[0059] In some embodiments, the camera device 102 can capture a color chart image by the first photosensitive module 106, and use the color chart image to train the color correction model. The standard color chart can be a multi-color standard color chart (such as a 24-color scale color chart). Taking a multi-color standard color chart as an example, the multi-color standard color chart can contain multiple square areas, each square having a different color. The first photosensitive module 106 can convert the spectral quantity of the color chart image into corresponding ambient light parameters. Alternatively, the camera device 102 can capture an environmental image of the test environment using the first photosensitive module 106, and convert the environmental image into ambient light parameters. The pre-processed ambient light parameters can include tristimulus value curves, spectral response functions, and / or the reflectance of the color chart. The reflectance of the color chart can be the reflectance of a standard 24-color chart, or it can be replaced with a regular rectangular wave, a custom curve, etc. The tristimulus value curve can be a tristimulus value curve under CIE 1931 or other specifications. Accordingly, the color correction model can be trained using color chart images from different light source environments, with the tristimulus curves, spectral response functions, and / or reflectance of the color chart images as the training set.

[0060] In step S2, as Figure 5As shown. Processor 104 can detect a region of interest A1 in the color-corrected image. Region of interest A1 can refer to a specified part of the human body or a non-specified part of the human body (e.g., the face region, hand region, or part of the human skin). Processor 104 can be a human detection machine learning model or a chip with a built-in human detection algorithm. The human detection machine learning model obtains and outputs the region of interest A1 based on the color-corrected image. Such human detection machine learning models include, but are not limited to, MediaPipe Pose Landmarker, YOLO, OpenPose, HRNet / HigherHRNET, CPN (Cascaded Pyramid Network), AlphaPose, PoseNet, MoveNet, BlazePose, OmniPose, HEViTPose (high-efficiency vision transformer for humanpose estimation), etc. After the human detection machine learning model identifies the human body, it can further label the region of interest A1. In some embodiments, the human detection machine learning model can also identify the region of interest A1 by detecting the human skin color.

[0061] In step S3, the processor 104 can obtain at least three different color light change signals—a first color light change signal, a second color light change signal, and a third color light change signal—from the pixels within the region of interest A1. In some embodiments, the second photosensitive module 108 can establish at least three different color light channels, such as a red light channel, a green light channel, and a blue light channel, within a single pixel of the captured target image TI. The processor 104 can detect data changes in the light signals of the red light channel, green light channel, and blue light channel (hereinafter referred to as RGB channels) within the region of interest A1. Here, the processor 104 can include red light signals, green light signals, and blue light signals within the region of interest A1.

[0062] In some embodiments, step S3 further includes: the processor 104 can calculate the RGB channels respectively according to a photoplethysmography (PPG) method to obtain the first color light change signal, the second color light change signal and the third color light change signal of the corresponding RGB channels.

[0063] In some embodiments, the first color light change signal includes a first color light DC component and a first color light AC component; the second color light change signal includes a second color light DC component and a second color light AC component; and the third color light change signal includes a third color light DC component and a third color light AC component. Specifically, the processor 104 can extract the DC and AC components of each color light channel using photovolume change recording. The DC component represents the average light intensity level that changes slowly over time after the light source irradiates the human body, formed by absorption and reflection by the skin, subcutaneous tissue, and non-pulsatile blood components. The AC component is the light intensity fluctuation formed by changes in arterial blood volume caused by heartbeat. Taking the RGB channel as an example, the first color light change signal is a red light change signal, from which the red light DC component and red light AC component can be obtained; the second color light change signal is a green light change signal, from which the green light DC component and green light AC component can be obtained; and the third color light change signal is a blue light change signal, from which the blue light DC component and blue light AC component can be obtained. It should be noted that the at least three different color light change signals are not limited to the aforementioned red, green and blue light, but can also be light of other different wavelengths.

[0064] Step S4 further includes: the processor 104 obtaining the light ratio parameter according to a light ratio estimation formula. The light ratio estimation formula is:

[0065]

[0066] RoR is the light ratio parameter; AC1 is the AC component of the first color light; DC1 is the DC component of the first color light; AC2 is the AC component of the second color light; DC2 is the DC component of the second color light; AC3 is the AC component of the third color light; DC3 is the DC component of the third color light; α, β, m, and n are constants associated with the second photosensitive module 108. Among them, α, β, m, and n are constants associated with the second photosensitive module 108 (linear or nonlinear laboratory constants). For example, when m and n are "1", this constant is a linear constant. These constants may vary depending on the type, adjustment, or setting of the second photosensitive module 108, and the aforementioned constants can be obtained by measurement using a laboratory ambient light source (standard light source).

[0067] In step S5, the processor 104 can obtain the blood oxygen saturation based on the light ratio parameter and a blood oxygen saturation formula. The blood oxygen saturation formula is:

[0068]

[0069] SpO2 is blood oxygen saturation (which can refer to pulse oxygen saturation); RoR is the light ratio parameter; A and B are constants. Among them, A and B are linear laboratory constants, which can vary depending on the type, adjustment, or setting of the second photosensitive module 108, and the aforementioned constants can be obtained by measurement in a laboratory ambient light source (standard light source).

[0070] Accordingly, the non-contact blood oxygen saturation measuring device 100 can directly calculate the light ratio parameter using light signals (e.g., RGB signals) of at least three wavelengths measurable from ambient light sources, and estimate blood oxygen saturation using this light ratio parameter. Furthermore, before estimating blood oxygen saturation, the processor 104 of the non-contact blood oxygen saturation measuring device 100 can perform a color correction mechanism. This ensures that the three-color light signals sensed by the second photosensitive module 108 under the current ambient light source can be corrected to be consistent with the laboratory ambient light source, so that the calculated light ratio parameter is not affected by changes in the ambient light source and thus does not amplify the error.

[0071] In some embodiments, such as Figure 1 As shown. The non-contact blood oxygen saturation measuring device 100 also includes a storage module 110 for storing information such as ambient light parameters and target image TI captured by the camera device 102. The storage module 110 can also store human body detection machine learning models, color correction models, light ratio estimation formulas, and blood oxygen saturation formulas for execution by the processor 104.

[0072] In some embodiments, such as Figure 6 As shown, the second photosensitive module 108 has multiple filter arrays 112, each including a first color light filter 114, a second color light filter 116, and a third color light filter 118. The processor 104 forms multiple optical channels based on these first color light filters 114, second color light filters 116, and third color light filters 118 to obtain red, green, and blue light signals. In some embodiments, the light signal can also be visible or invisible light of other wavelengths, depending on the wavelength filtering range set by each filter (114, 116, 118). For example, the first color light filter 114 allows only red light Lr to pass through; the second color light filter 116 allows only green light Lg to pass through; and the third color light filter 118 allows only blue light Lb to pass through.

[0073] In summary, in some embodiments of the non-contact blood oxygen saturation measuring device 100, the first photosensitive module 106 captures ambient light parameters and the second photosensitive module 108 captures the target image TI to estimate blood oxygen saturation. Specifically, the processor 104 can obtain a prediction correction coefficient based on the ambient light parameters and use the prediction correction coefficient to perform a color correction mechanism on the target image TI, so that the color of the image captured by the camera device 102 under the current ambient light source is consistent with the color of the laboratory ambient light source. Therefore, when the processor 104 estimates blood oxygen saturation using the color-corrected image, it can avoid calculation errors in the estimated blood oxygen saturation caused by differences between the ambient light source and the laboratory ambient light source. In other words, the non-contact blood oxygen saturation measuring device 100 can, under varying and arbitrary everyday ambient light conditions, obtain accurate light ratio parameters by using signals (such as ambient light parameters and target image TI) captured by the first photosensitive module 106 and the second photosensitive module 108 and the calculations of the processor 104, eliminating the influence of environmental noise. This enables real-time estimation of blood oxygen saturation without interference from ambient light. Accordingly, the non-contact blood oxygen saturation measuring device 100 and its measurement and calibration method S in some embodiments can improve the limitations of traditional contact measurement and solve the technical bottleneck of existing non-contact technologies being unable to resist ambient light interference.

[0074] Of course, the present invention may have other various embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding changes and modifications according to the present invention, but these corresponding changes and modifications should all fall within the protection scope of the claims of the present invention.

Claims

1. A non-contact blood oxygen saturation measuring device, characterized in that, Include: A camera device, comprising: A first photosensitive module is used to capture an ambient light parameter; and A second photosensitive module for capturing a target image; and A processor for performing the following steps: A color-corrected image is obtained based on the target image and the ambient light parameters; A region of interest was detected in the color-corrected image; Based on the region of interest, a first color light change signal, a second color light change signal, and a third color light change signal of multiple different colors of light are obtained; A light ratio parameter is obtained based on the first color light change signal, the second color light change signal, and the third color light change signal; as well as Blood oxygen saturation is obtained based on this light ratio parameter.

2. The non-contact blood oxygen saturation measuring device as described in claim 1, characterized in that, The processor is used to execute a color correction model, with the ambient light parameter input into the color correction model to obtain a prediction correction coefficient, and the processor obtains the color-corrected image based on the target image and the prediction correction coefficient.

3. The non-contact blood oxygen saturation measuring device as described in claim 2, characterized in that, The color correction model is trained with a laboratory light source and multiple control ambient light sources to output a predicted correction coefficient, which is either a color matrix gain coefficient or a three primary color gain coefficient.

4. The non-contact blood oxygen saturation measuring device as described in claim 3, characterized in that, The color correction model is a convolutional neural network (CNN) model.

5. The non-contact blood oxygen saturation measuring device as described in claim 1, characterized in that, The first color light change signal includes a first color light DC component and a first color light AC component; the second color light change signal includes a second color light DC component and a second color light AC component; and the third color light change signal includes a third color light DC component and a third color light AC component.

6. The non-contact blood oxygen saturation measuring device as described in claim 5, characterized in that, The processor is used to obtain the light ratio parameter according to a light ratio estimation formula, which is as follows: Wherein, RoR is the light ratio parameter; AC1 is the first color light AC component; DC1 is the first color light DC component; AC2 is the second color light AC component; DC2 is the second color light DC component; AC3 is the third color light AC component; DC3 is the third color light DC component; α, β, m, and n are constants associated with the second photosensitive module.

7. The non-contact blood oxygen saturation measuring device as described in claim 5 or 6, characterized in that, The region of interest includes a red light signal, a green light signal, and a blue light signal; the processor obtains a first color change signal, a second color change signal, and a third color change signal corresponding to the red light signal, the green light signal, and the blue light signal.

8. The non-contact blood oxygen saturation measuring device as described in claim 7, characterized in that, The first photosensitive module has multiple filter arrays, each including a first color light filter, a second color light filter, and a third color light filter; the processor obtains the red light signal, the green light signal, and the blue light signal based on the first color light filter, the second color light filter, and the third color light filter, respectively.

9. The non-contact blood oxygen saturation measuring device as described in claim 1, characterized in that, The first photosensitive module is a multispectral sensor, and the second photosensitive module is an RGB photosensitive module.

10. A non-contact method for measuring and correcting blood oxygen saturation, characterized in that, A non-contact blood oxygenation measurement and correction method is suitable for a processor to correct a target image and to estimate a blood oxygen saturation based on the corrected target image. This method includes: Capture the target image and an ambient light parameter; A color-corrected image is obtained based on the target image and the ambient light parameters; A region of interest was detected in the color-corrected image; Based on the region of interest, a first color light change signal, a second color light change signal, and a third color light change signal of multiple different colors of light are obtained; A light ratio parameter is obtained based on the first color light change signal, the second color light change signal, and the third color light change signal; as well as The blood oxygen saturation is obtained based on this light ratio parameter.

11. The non-contact blood oxygen saturation measurement and correction method as described in claim 10, characterized in that, The step of acquiring the target image and the ambient light parameters also includes: A color correction model is executed, with the ambient light parameter input to the color correction model to obtain a predicted correction coefficient; and The color-corrected image is obtained based on the target image and the prediction correction coefficient.

Images