Non-contact blood oxygen monitoring method and system based on mobile phone direct current component calibration

Abstract

This application relates to the field of medical device technology, specifically proposing a non-contact blood oxygen monitoring method and system based on mobile phone DC component calibration. The method includes: acquiring the hand photoplethysmography (PPG) signal of the target user's hand using a constructed hand-reflective optical path and extracting the hand DC component and hand pulsation component; acquiring the target user's first facial video information, acquiring the target user's facial PPG signal and extracting the facial DC component; constructing a blood oxygen calibration model based on the hand DC component, hand pulsation component, and facial pulsation component; acquiring the target user's second facial video information, extracting the DC component from the second facial video information, and inputting it into the blood oxygen calibration model to obtain the target user's current blood oxygen information. This application improves the accuracy of continuous monitoring using only the front-facing camera when applying mobile phone cameras for daily blood oxygen monitoring by utilizing the collaborative work of front and rear cameras.

Description

Technical Field

[0001] This invention relates to the field of medical device technology, specifically to a non-contact blood oxygen monitoring method and system based on mobile phone DC component calibration. Background Technology

[0002] Blood oxygen saturation ( (IV) is an important vital sign for assessing human respiratory function and oxygen supply capacity. With the widespread use of smartphones and the continuous improvement of their camera performance, non-contact measurement using smartphone cameras has emerged. The monitoring method involves acquiring facial video signals through a front-facing camera, extracting the pulsation component (AC component) from them, and indirectly estimating blood oxygen levels.

[0003] First, although the above method achieves non-contact blood oxygen estimation, the AC component has a low signal-to-noise ratio under visible light conditions and is easily affected by interference factors such as ambient lighting, head micro-movements, and skin color differences.

[0004] Secondly, professional blood oxygen monitoring equipment often uses dual-wavelength light sources in the red and near-infrared (NIR) bands for reflective detection to improve the ability to distinguish between oxygenated and deoxygenated hemoglobin in the blood. However, the front-facing camera of ordinary smartphones usually does not have the ability to collect NIR band data, which limits its accuracy and generalization ability in blood oxygen estimation.

[0005] Therefore, given the shortcomings of traditional technologies, there is an urgent need to address the problems of low personal adaptability, high susceptibility to environmental influences, and poor accuracy in the field of non-contact blood oxygen monitoring using mobile phone cameras. Summary of the Invention

[0006] The primary objective of this application is to address at least one of the aforementioned problems by providing a non-contact blood oxygen monitoring method and system based on mobile phone DC component calibration.

[0007] To achieve the various objectives of this application, the following technical solution is adopted:

[0008] A non-contact blood oxygen monitoring method based on mobile phone DC component calibration, provided for one of the purposes of this application, includes the following steps:

[0009] Based on the constructed hand reflective optical path, the target user's hand photoplethysmography (PPG) signal is acquired and the DC component and pulsation component of the hand are extracted; the target user's first facial video information is acquired, and the target user's facial PPG signal is acquired and the facial DC component is extracted; a blood oxygen calibration model is constructed based on the hand DC component, hand pulsation component, and facial pulsation component; the target user's second facial video information is acquired, the DC component in the second facial video information is extracted and input into the blood oxygen calibration model to obtain the target user's current blood oxygen information.

[0010] Optionally, the step of acquiring the photoplethysmography (PPG) signal of the target user's hand based on the constructed hand-reflective optical path and extracting the DC component and pulsation component of the hand includes the following steps: acquiring RGB video signals containing the hand-reflective optical path constructed when the target user's hand covers the camera and performing sliding window processing to obtain video data for each channel; using a bandpass filter to extract the pulsation component of the video data for each channel respectively, and using a low-pass filter to remove high-frequency noise from the video data for each channel respectively, and obtaining the filtered mean as the DC component of the hand for the corresponding time window.

[0011] Optionally, the step of acquiring the first facial video information of the target user, collecting the facial photoplethysmography (PPG) signal of the target user and extracting the facial DC component includes the following steps: acquiring the first facial video information containing the target user's face and performing sliding window processing, and using a low-pass filter to extract the facial DC component within each time window.

[0012] Optionally, the step of constructing a blood oxygen calibration model based on the hand DC component, hand pulsation component, and facial pulsation component includes the following steps: obtaining a standard blood oxygen saturation based on the relative amplitude ratio of the hand pulsation component and the hand DC component of each channel through a linear mapping relationship between the relative amplitude ratio and blood oxygen saturation; obtaining an estimated blood oxygen saturation based on the relative ratio of the facial DC component at different wavelengths through a linear mapping relationship between the relative ratio and blood oxygen saturation; calculating the deviation between the standard blood oxygen saturation and the estimated blood oxygen saturation to obtain a predicted deviation value; constructing a linear translation function based on the predicted deviation value; correcting the mapping function of the estimated blood oxygen saturation through the linear translation function so that the fitting result of the mapping function of the estimated blood oxygen saturation matches the target user; and synchronizing the corrected mapping function of the blood oxygen saturation to the blood oxygen calibration model.

[0013] Optionally, the step of acquiring the hand photoplethysmography (PPG) signal of the target user based on the constructed hand reflective optical path and extracting the DC component and pulsation component of the hand; and acquiring the first facial video information of the target user, after acquiring the facial PPG signal of the target user and extracting the facial DC component, includes the following steps: aligning the hand PPG signal and the facial PPG signal by timestamps to achieve data time alignment and physiological state consistency control.

[0014] Optionally, after constructing the blood oxygen calibration model based on the hand DC component, hand pulsation component, and facial pulsation component, the following steps are included: creating a target user blood oxygen information database based on the first facial video information, mapping the blood oxygen calibration model to the target user blood oxygen information database, and calling the latest blood oxygen calibration model in the target user blood oxygen information database according to the correspondence between the second facial video information and the first facial video information.

[0015] Optionally, the step of acquiring the first facial video information of the target user, collecting the facial photoplethysmography (PPG) signal of the target user, and extracting the facial pulsation component further includes the following steps: correcting the face in the first facial video information, dividing the face region into multiple regions of interest based on multiple facial muscle key points; filtering the corresponding facial muscle key points according to the pre-set regions of interest, and selecting the high-reflectivity region for collecting the facial PPG signal; collecting the facial PPG signal of the high-reflectivity region, and extracting the facial pulsation component from it.

[0016] On the other hand, a non-contact blood oxygen monitoring system based on mobile phone DC component calibration, provided to meet one of the purposes of this application, includes:

[0017] The acquisition module is used to acquire the photoplethysmography (PPG) signal of the target user's hand based on the constructed reflective optical path of the target user's hand; and to acquire the first facial video information of the target user, acquire the facial PPG signal of the target user and extract the facial pulsation component.

[0018] A construction module is used to construct a blood oxygen calibration model based on the DC component of the hand, the pulsation component of the hand, and the pulsation component of the face;

[0019] The analysis module is used to acquire second facial video information, extract the DC component from the second facial video information and input it into the blood oxygen calibration model to obtain the current blood oxygen information of the target user.

[0020] On another note, a non-contact blood oxygen monitoring device based on mobile phone DC component calibration, provided to meet one of the purposes of this application, includes a central processing unit and a memory. The central processing unit is used to call and run a computer program stored in the memory to execute the steps of the non-contact blood oxygen monitoring method based on mobile phone DC component calibration described in this application.

[0021] In another aspect, a computer-readable storage medium is provided to suit one of the purposes of this application, the computer-readable storage medium storing computer-executable instructions for causing a computer to perform a non-contact blood oxygen monitoring method based on mobile phone DC component calibration as disclosed in any of the first aspects of the present invention.

[0022] The technical solution of this application has many advantages, including but not limited to the following aspects:

[0023] This application first acquires the photoplethysmography (PPG) signal of the target user's hand using a constructed reflective optical path before initial use, extracting the DC component and pulsation component of the hand to effectively remove interference from ambient light changes and achieve high-quality PPG signal acquisition. Simultaneously, it acquires first facial video information in real time, completing the synchronous acquisition of non-contact facial PPG signals. It fully utilizes the high-quality fingertip contact PPG signal acquired by the rear camera as a reference gold standard, thereby improving the accuracy of blood oxygen monitoring based solely on facial PPG signals.

[0024] Secondly, a blood oxygen calibration model is constructed based on the DC component of the hand, the pulsation component of the hand, and the pulsation component of the face. After the initial calibration is completed, the DC component of the facial DC light reflection signal collected by the front camera can be individually calibrated. This allows users to achieve continuous blood oxygen monitoring without finger contact and relying solely on the front camera after the initial calibration. This avoids the influence of motion, lighting, and skin color on the pulsation component in actual acquisition, while leveraging the stability advantage of the DC component in steady-state trend tracking. Attached Figure Description

[0025] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0026] Figure 1 This is a schematic diagram illustrating an exemplary data processing flow of this application;

[0027] Figure 2 This is a flowchart of one embodiment of the non-contact blood oxygen monitoring method based on mobile phone DC component calibration according to this application;

[0028] Figure 3This is a schematic diagram of the non-contact blood oxygen monitoring system based on mobile phone DC component calibration used in this application. Detailed Implementation

[0029] Before detailing the specific embodiments of the technical solution of this application, we will first disclose the application scenarios suitable for supporting the architecture and early warning method of the photovoltaic equipment in the technical solution of this application.

[0030] With the widespread adoption of smartphones and the continuous improvement of their camera performance, researchers in related fields have begun to explore the use of smartphone cameras for non-contact... Monitoring methods in this field are mostly based on remote photoplethysmography (rPPG) technology. rPPG technology uses subtle color changes in the skin caused by ambient light to monitor blood flow. When the heart beats, blood flow causes changes in the intensity and color of light reflected from the skin. These changes are captured by a high-definition camera and converted into a photoplethysmography (PPG) pulse wave signal containing both AC and DC components. The AC component is used to indirectly estimate the pulse wave. However, the AC component has a low signal-to-noise ratio under visible light conditions and is easily affected by factors such as ambient lighting, head movements, and skin color differences. In contrast, the DC component in the light signal reflects the average light absorption at the tissue level and is related to tissue oxygen saturation. Related to this, it can be indirectly used The DC component appears as a stable light reflection background in facial videos, which is theoretically more suitable for monitoring trend-based physiological indicators. It is not sensitive to motion interference and has a high signal-to-noise ratio. However, the absolute value of the DC component is easily affected by skin color (differences in melanin absorption), spectral changes, and the camera's automatic exposure mechanism. Therefore, it is difficult to use it directly for blood oxygen estimation and still needs to rely on individualized calibration methods to improve accuracy.

[0031] It is evident that non-contact blood oxygen monitoring technology on smartphone platforms still faces the following key challenges:

[0032] 1. Rear camera relies on contact monitoring: Although the quality of fingertip contact photoplethysmography (PPG) signals is high, its operation method restricts user freedom. Each time monitoring is performed, the finger must be completely covered by the camera, making it difficult to achieve non-contact monitoring.

[0033] 2. Low signal-to-noise ratio of front-facing camera: rPPG technology relies on AC components, which are easily affected by environmental changes, movement and individual skin color, resulting in poor signal stability and limited generalization ability.

[0034] Third, deep learning methods are computationally limited: they require large-scale labeled data, model training and inference are complex, and they are difficult to deploy on low-power devices such as mobile phones.

[0035] Fourth, professional NIR systems are expensive: Although NIR imaging has high precision and stability in the medical field, its multi-wavelength light-emitting modules and photosensitive components usually need to be specially customized, resulting in high costs and large equipment, making it difficult to popularize and apply in mobile platforms or everyday scenarios.

[0036] Existing technologies are mostly limited to using a single camera to collect signals from a single body part, or rely on contact operation, failing to fully explore the potential of dual-camera structures in blood oxygen estimation. This application utilizes the front and rear cameras of a smartphone in tandem, using the high-quality fingertip photoplethysmography (FP-P) signal acquired by the rear camera as the gold standard. This allows for individualized calibration and correction of the DC component acquired by the front camera, enabling users to achieve continuous blood oxygen monitoring without finger contact, relying solely on the front camera, after initial calibration.

[0037] The technical solution of this application is applicable to the field of blood oxygen monitoring medical device technology, and is particularly applicable to the scenario of monitoring blood oxygen with a smartphone camera. In this context, the technical solution of this application can be applied in a typical smartphone device, which includes a control board, a front camera and a rear camera.

[0038] The control board, as a control element, is responsible for receiving video data from the front and rear cameras, analyzing the video data to obtain the AC and DC components in the photoplethysmography (PPG) signal, and constructing a blood oxygen calibration model that can be used for calibration to correct subsequent monitoring behavior. This fully utilizes the native hardware of smartphones, greatly improves user acceptance and universality, and reduces system costs.

[0039] Among them, the front-facing camera, as a facial information acquisition element, is responsible for non-contact continuous monitoring of the DC component of the facial photoplethysmography (rPPG) signal; the rear-facing camera, as a hand information acquisition element, acquires high-quality photoplethysmography (rPPG) signals from the fingertips for individualized model calibration; this collaborative framework breaks through the limitation of low accuracy in traditional single-camera non-contact rPPG methods and achieves complementary advantages of dual channels.

[0040] It is understood that the embodiments of this application are not limited to front and rear cameras. Given that a smartphone has multiple cameras, the multiple cameras of the phone can be used together to collect image data from different angles and bands. The fusion algorithm can be used to improve signal quality and anti-interference ability, thereby expanding the scope of application.

[0041] In some embodiments, the technical solutions of this application can also be applied to other smart terminal devices, such as smartwatches, smart glasses, monitoring instruments, etc., thereby realizing more diverse non-contact blood oxygen monitoring scenarios.

[0042] Furthermore, to address the aforementioned technical issues, this application proposes a non-contact blood oxygen monitoring method based on mobile phone DC component calibration, aiming to establish a method that can continuously monitor blood oxygen levels in real-life scenarios without rear-facing contact, relying solely on the front-facing camera. This enhances the anti-interference capabilities and accuracy of non-contact blood oxygen monitoring via smartphone cameras.

[0043] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments.

[0044] The specific embodiments described below can be combined with each other, and the same or similar concepts or processes will not be described again in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.

[0045] See Figure 2 This application discloses a non-contact blood oxygen monitoring method based on mobile phone DC component calibration. In its typical blood oxygen monitoring embodiment, it includes the following steps:

[0046] 101: Based on the constructed target user's hand reflective optical path, acquire the target user's hand photoplethysmography (PPG) signal; and acquire the target user's first facial video information, acquire the target user's facial PPG signal and extract the facial pulsation component.

[0047] The construction of the target user's hand reflective optical path can be configured by outputting instructions to instruct the target user to cover the smartphone's rear camera and flash with their finger upon first use, so that the light from the flash illuminates the finger. At the same time, the adjacent rear camera captures video information of the finger being illuminated, analyzes the subtle color changes in the video information, and collects the hand photoplethysmography signal containing the hand's pulsation component and DC component.

[0048] Smartphone cameras are typically able to capture video information in RGB format, which includes three primary color channels and can be divided into multiple video information of different wavelengths, as well as multiple hand photoplethysmography (PPG) signals of different wavelengths. The Ratio-of-Ratios (RoR) method is used to combine the pulsation components and DC components of different wavelengths to estimate blood oxygen saturation. The video information of different wavelengths can be red light or green light.

[0049] In application, the system processes the RGB video signals acquired by the front and rear cameras respectively, extracts the hand pulsation component, the hand DC component, and the face DC component, and constructs RoR features that reflect the optical absorption characteristics of different bands.

[0050] In practice, RGB video signals containing the reflective optical path of the hand when the target user's hand covers the camera are acquired and processed by sliding window to obtain video data for each channel. A bandpass filter is used to extract the pulsating component of the hand photoplethysmography (PPG) signal of each channel video data, and a low-pass filter is used to remove high-frequency noise from each channel video data. The filtered mean is used as the DC component of the hand PPG signal for the corresponding time window. Then, based on the relative amplitude ratio of the pulsating component and the DC component of the hand PPG signal of each channel, the standard blood oxygen saturation is obtained through the linear mapping relationship between the relative amplitude ratio and blood oxygen saturation.

[0051] By completely covering the rear camera and fill light with your finger, you can prevent the rear camera from receiving ambient light, thereby eliminating interference caused by changes in ambient light and achieving high-quality acquisition of hand photoplethysmography (PPG) signals.

[0052] Sliding window processing refers to dividing the RGB video signal into several fixed-size time windows and performing data analysis within each time window. By sliding the time window, statistical indicators of data within a certain time period can be calculated in real time, which can effectively improve the accuracy of real-time blood oxygen monitoring.

[0053] First, a fourth-order Butterworth bandpass filter with a frequency range of 0.6Hz to 3Hz is used to extract the pulsation components of each color channel. Then, a fourth-order Butterworth low-pass filter with a cutoff frequency of 0.6Hz is used to remove high-frequency noise from the facial photoplethysmography (PPG) signal, making the subsequently calculated signal data purer and improving the anti-interference capability of the calculation process. Subsequently, a feature curve reflecting the optical absorption characteristics of different bands is constructed. :

[0054] ;

[0055] in, The ratio of AC is the value of AC. This represents the pulsating component of the red channel in the hand photoplethysmography (PPG) signal. This represents the DC component of the red channel in the hand photoplethysmography (PPG) signal. The pulsating component of the green channel in the photoplethysmography (PPG) signal of the hand. This refers to the DC component of the green channel in the hand photoplethysmography (PPG) signal.

[0056] Through features It can calculate a relatively accurate standard blood oxygen saturation, which serves as the gold standard for individual calibration of the front-facing camera DC model.

[0057] First facial video information containing the target user's face is acquired and processed using a sliding window method. A low-pass filter is used to extract the DC component of the facial photoplethysmography (PPG) signal within each time window. Then, based on the relative ratios of the DC components of the facial PPG signals at different wavelengths, an estimated blood oxygen saturation is obtained through a linear mapping relationship between these ratios and blood oxygen saturation. The sliding window processing steps for the first facial video are the same as those for the hand PPG signal described above.

[0058] In this embodiment, different wavelengths, including RGB three channels, are used. By combining a sliding window averaging algorithm to smooth the time series, a stable DC component in the target user's facial region can be extracted. Then, the RoR principle is used to calculate the DC components in different wavelength bands. Linear fitting curve of blood oxygen under the following conditions:

[0059] ;

[0060] in, This represents the ratio of the pure DC component in the facial photoplethysmography (PPG) signal. This represents the DC component of the red channel in the facial photoplethysmography (PPG) signal. This refers to the DC component of the green channel in the facial photoplethysmography signal.

[0061] Features extracted from the rear camera It can fit the obtained linear blood oxygen fitting curve through pre-calibration, thereby obtaining an estimated blood oxygen saturation.

[0062] In some implementations, illumination normalization and skin color mapping mechanisms can be introduced simultaneously to reduce the impact of external environmental changes on the stability of the DC component in the facial photoplethysmography signal.

[0063] 102: Construct a blood oxygen calibration model based on the DC component of the hand, the pulsation component of the hand, and the pulsation component of the face.

[0064] The blood oxygen calibration model can quickly convert a general model into an individual module based on the target user's standard blood oxygen saturation calibrated by the rear camera. Before modeling, dual-camera collaborative recording achieves data temporal alignment and physiological state consistency control, providing a reliable foundation for subsequent modeling. Specifically, the hand photoplethysmography (PPG) wave signals and facial PPG wave signals are aligned using timestamps to achieve data temporal alignment and physiological state consistency control.

[0065] During the initial use, users need to perform a brief breath-holding experiment to induce controllable changes in blood oxygen saturation over a wider range. The rear camera then collects high-quality photoplethysmography (PPG) signals from the fingertip to estimate the current true blood oxygen saturation. Horizontal distance is used as the gold standard. The system synchronously records the corresponding time period of the front-facing camera. Output the results. Then, compare the data based on this time period. Model estimation Value and Post-True By analyzing the deviations between values, a linear translation function is constructed to individually tune the RoR-DC model, correcting the baseline of its predicted values ​​and enabling a rapid conversion from a general model to an individualized model. This process does not change the RoR feature construction method, but only adjusts the mapping benchmark of its corresponding output, improving estimation accuracy. After calibration, this individualized blood oxygen calibration model will be saved to the local device for subsequent continuous monitoring.

[0066] In practice, the standard blood oxygen saturation and the estimated blood oxygen saturation are compared to obtain the prediction deviation value between the two, and a linear translation function is constructed based on the prediction deviation value. The mapping function of the estimated blood oxygen saturation is then corrected using the linear translation function so that the fitting result of the mapping function matches the target user. The corrected mapping function of blood oxygen saturation is then synchronized to the blood oxygen calibration model. Specifically, the user performs a brief breath-holding operation under system guidance, while the front-facing camera simultaneously captures data. Features, combined with those extracted from the rear camera Features were identified, and standard blood oxygen saturation was estimated based on a priori physiological models. The established linear mapping relationship is as follows:

[0067] ;

[0068] in, Both b and are standard blood oxygen saturation calibration coefficients obtained through fitting.

[0069] Because it uses a rear-facing camera and a contact-based monitoring method, it avoids interference from ambient light and provides a relatively stable measurement spectrum, thus enabling the acquisition of accurate standard blood oxygen saturation. Meanwhile, based on the corresponding time period of the front-facing camera Features, using linear fitting to construct estimated blood oxygen saturation under DC component Mapping function:

[0070] ;

[0071] in, and All of these are estimated blood oxygen saturation calibration coefficients obtained through fitting.

[0072] Standard blood oxygen saturation Linear mapping function and estimation of blood oxygen saturation By comparing and obtaining the prediction deviation value, a linear translation function is constructed to estimate blood oxygen saturation. The mapping function is close to the standard blood oxygen saturation. The mapping function allows users to directly utilize the DC component of the facial photoplethysmography signal obtained from the front camera to achieve more accurate blood oxygen monitoring. After modeling is completed, the blood oxygen calibration model based on the current spectral environment is saved for subsequent continuous estimation.

[0073] In some implementations, the linear translation calibration model can be replaced with a more complex model such as nonlinear regression or neural networks for individualized mapping, thereby improving the fitting accuracy.

[0074] 103: Obtain the second facial video information, extract the DC component from the second facial video information and input it into the blood oxygen calibration model to obtain the current blood oxygen information of the target user.

[0075] Once the modeling is complete, users no longer need to touch the rear camera during daily use. The system only needs to have the user's face directly facing the front camera to automatically capture second-face video information, extract the DC component, and input it into the established individual blood oxygen calibration model for real-time estimation. level.

[0076] In some implementations, a target user blood oxygen information database is created based on the first facial information, and the blood oxygen calibration model is mapped to the target user blood oxygen information database. The latest blood oxygen calibration model in the target user blood oxygen information database is then called based on the correspondence between the second facial video information and the first facial information. When a user uses the front-facing camera for daily blood oxygen monitoring, the acquired second facial video information is matched with the same first facial information, and the corresponding blood oxygen calibration model is called from the target user blood oxygen information database, ensuring that different users can accurately use the personalized model corresponding to themselves. It is understood that the target user blood oxygen information database also includes the target user's historical blood oxygen data.

[0077] This method offers contactless, continuous monitoring capabilities, making it suitable for seamless health tracking in everyday scenarios. The system also supports periodic recalibration or dynamic model adjustment based on DC component variation trends, enhancing adaptability and stability.

[0078] In specific implementation, the face in the first facial video information is corrected, and the face region is divided into multiple regions of interest based on multiple facial muscle key points; the corresponding facial muscle key points are selected according to the pre-set regions of interest, and a high-reflectivity region for collecting facial photoplethysmography (PPG) signals is selected, the facial PPG signals of the high-reflectivity region are collected, and the facial pulsation component is extracted from them.

[0079] Among them, facial correction is performed by correcting the face to be facing forward to improve the accuracy of recognition. When implementing this scheme, the region of interest can be selected based on prior experience from locations with high reflectivity, such as the forehead and cheekbones, thereby increasing the amount of facial photoplethysmography (PPG) signal data acquired and providing a more stable and reliable data foundation for subsequent blood oxygen monitoring.

[0080] Finally, as Figure 1 As shown, the workflow of this application is as follows: upon initial use, a hand photoplethysmography (PPG) signal is simultaneously acquired via a rear camera and a face PPG signal is acquired via a front camera. The pulsation and DC components of each color channel are extracted from the hand PPG signal. A standard blood oxygen saturation mapping function is constructed using the pulsation and DC components from the hand PPG signal. The DC components of each color channel are extracted from the face PPG signal. An estimated blood oxygen saturation mapping function is constructed using the DC components from the face PPG signal. A linear translation function is used to bring the estimated blood oxygen saturation mapping function closer to the standard blood oxygen saturation mapping function. The estimated blood oxygen saturation mapping function is then corrected, thereby obtaining a blood oxygen calibration model that can accurately monitor blood oxygen saturation using only the DC component of the face PPG signal.

[0081] The unique technical advantage of this application lies in the fact that the method is non-contact, sustainable, low-cost, and easy to promote. It takes into account both signal quality and user convenience, and is applicable to multiple scenarios such as home monitoring, mobile health management, and telemedicine. It has good clinical application prospects and industrial transformation potential.

[0082] Please see Figure 3 According to one aspect of this application, a non-contact blood oxygen monitoring system based on mobile phone DC component calibration is provided. The system includes: an acquisition module 201, used to acquire the hand photoplethysmography (PPG) signal of the target user based on a constructed hand reflective optical path; and to acquire first facial video information of the target user, acquire the facial PPG signal of the target user, and extract the facial pulsation component; a construction module 202, used to construct a blood oxygen calibration model based on the hand DC component, hand pulsation component, and facial pulsation component; and an analysis module 203, used to acquire second facial video information, extract the DC component from the second facial video information, and input it into the blood oxygen calibration model to obtain the current blood oxygen information of the target user.

[0083] Based on any embodiment of the system in this application, the system of this application further includes: a hand signal acquisition module, configured to acquire RGB video signals including the hand reflective optical path constructed when the target user's hand covers the camera and perform sliding window processing to obtain video data for each channel; using a bandpass filter to extract the pulsating component of the hand photoplethysmography (PPG) signal of each channel video data respectively, and using a low-pass filter to remove high-frequency noise from each channel video data respectively, to obtain the filtered mean as the DC component of the hand PPG signal for the corresponding time window.

[0084] Based on any embodiment of the system in this application, the system of this application further includes: a facial signal acquisition module, configured to acquire first facial video information containing the face of the target user and perform sliding window processing, and to extract the DC component of the facial photoplethysmography signal in each time window using a low-pass filter.

[0085] Based on any embodiment of the system in this application, the system further includes: a result correction module, configured to obtain standard blood oxygen saturation based on the relative amplitude ratio of the pulsating component and the DC component of the hand photoplethysmography (PPG) signal of each channel, through a linear mapping relationship between the relative amplitude ratio and blood oxygen saturation; to obtain estimated blood oxygen saturation based on the relative ratio of the DC component of the hand PPG signal at different wavelengths, through a linear mapping relationship between the relative ratio and blood oxygen saturation; to compare the standard blood oxygen saturation and the estimated blood oxygen saturation and obtain a prediction deviation value between the two, and to construct a linear translation function based on the prediction deviation value; to correct the mapping function of the estimated blood oxygen saturation through the linear translation function, so that the fitting result of the mapping function of the estimated blood oxygen saturation matches the target user, and to synchronize the corrected mapping function of blood oxygen saturation to the blood oxygen calibration model.

[0086] Based on any embodiment of the system in this application, the system of this application further includes: creating a target user blood oxygen information database based on the first facial information, mapping the blood oxygen calibration model to the target user blood oxygen information database, and calling the latest blood oxygen calibration model in the target user blood oxygen information database according to the correspondence between the second facial video information and the first facial information.

[0087] Based on any embodiment of the system in this application, the system of this application further includes: a facial information optimization module, configured to correct the face in the first facial video information, divide the face region into multiple regions of interest according to multiple facial muscle key points; filter the corresponding facial muscle key points according to the preset regions of interest, and select the high reflectivity region for collecting facial photoplethysmography signals.

[0088] Another embodiment of this application provides a non-contact pulse oximetry monitoring device based on mobile phone DC component calibration. This device includes a processor, a computer-readable storage medium, a memory, and a network interface connected via a system bus. The computer-readable, non-volatile storage medium stores an operating system, a database, and computer-readable instructions. The database may store information sequences. When the computer-readable instructions are executed by the processor, the processor can implement a non-contact pulse oximetry monitoring method based on mobile phone DC component calibration.

[0089] The processor of this non-contact pulse oximetry monitoring device based on mobile phone DC component calibration provides computing and control capabilities, supporting the operation of the entire device. The device's memory can store computer-readable instructions, which, when executed by the processor, cause the processor to perform the non-contact pulse oximetry monitoring method based on mobile phone DC component calibration of this application. The network interface of the device is used for communication with a terminal.

[0090] In this embodiment, the processor is used to execute... Figure 3 The system defines the specific functions of each module, and the memory stores the program code and various data required to execute these modules or submodules. The network interface is used to enable data transmission between user terminals or servers.

[0091] The non-volatile readable storage medium in this embodiment stores the program code and data required to execute all modules in the non-contact blood oxygen monitoring system based on mobile phone DC component calibration of this application. The server can call the program code and data of the server to execute the functions of all modules.

[0092] This application also provides a non-volatile readable storage medium storing computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps of the non-contact blood oxygen monitoring method based on mobile phone DC component calibration according to any embodiment of this application.

[0093] This application also provides a computer program product, including a computer program / instructions that, when executed by one or more processors, implement the steps of the method described in any embodiment of this application.

[0094] In summary, this application provides a solution for non-contact blood oxygen monitoring based on a mobile phone camera that avoids environmental interference. Building upon existing non-contact blood oxygen monitoring technologies, this invention proposes several key innovations aimed at improving the accuracy, stability, and ease of use of blood oxygen estimation. The main differences between this invention and traditional technologies are as follows:

[0095] 1. Front and Rear Camera Collaborative Acquisition Architecture: This paper proposes for the first time a collaborative working mode utilizing the front and rear cameras built into a smartphone. The front camera acquires the DC component for non-contact continuous monitoring, while the rear camera obtains high-quality photoplethysmography (rPPG) signals from the fingertip for individualized model calibration. This collaborative framework breaks through the limitation of low accuracy in traditional single-camera non-contact rPPG methods, achieving complementary advantages of dual channels.

[0096] 2. RoR-DC-based signal construction and calibration strategy: This invention proposes to construct the DC component characteristics of the front-facing camera based on the RoR principle, and further designs a RoR-DC linear mapping and offset calibration mechanism for the rear-facing camera. By using the DC model as a reference gold standard, baseline shifting was performed, effectively eliminating systematic biases caused by individual skin differences and camera light sensitivity differences.

[0097] 3. No external sensors required, completely non-contact design: Unlike methods that require additional wearable devices or dedicated imaging hardware (such as NIR systems or multispectral modules), this invention relies solely on the native hardware of a smartphone to achieve high-precision blood oxygen estimation, significantly improving its user acceptance and universality, and reducing system costs.

[0098] 4. Model update mechanism adapted to dynamic environment: The system supports local deployment and lightweight updates of the model, and can periodically recalibrate or dynamically adjust the model according to the trend of DC component changes, thereby enhancing adaptability and stability.

[0099] These innovations together form a complete, practical, low-cost, and contactless blood oxygen estimation solution applicable to existing mobile phone platforms, providing a new path for remote health monitoring and family health management.

Claims

1. A non-contact blood oxygen monitoring method based on mobile phone DC component calibration, characterized in that, Includes the following steps: Based on the constructed target user's hand reflective optical path, the target user's hand photoplethysmography (PPG) signal is acquired and the DC component and hand pulsation component of the hand are extracted; and the first facial video information of the target user is acquired, and the target user's facial PPG signal is acquired and the facial DC component is extracted. A blood oxygen calibration model is constructed based on the DC component of the hand, the pulsation component of the hand, and the pulsation component of the face. Acquire the second facial video information of the target user, extract the DC component from the second facial video information and input it into the blood oxygen calibration model to obtain the current blood oxygen information of the target user; The steps for constructing a blood oxygen calibration model based on the hand DC component, hand pulsation component, and facial pulsation component include: obtaining a standard blood oxygen saturation based on the relative amplitude ratio of the hand pulsation component and the hand DC component in each channel, through a linear mapping relationship between the relative amplitude ratio and blood oxygen saturation; obtaining an estimated blood oxygen saturation based on the relative ratio of the facial DC component at different wavelengths, through a linear mapping relationship between the relative ratio and blood oxygen saturation; calculating the deviation between the standard blood oxygen saturation and the estimated blood oxygen saturation to obtain a predicted deviation value, constructing a linear translation function based on the predicted deviation value; correcting the mapping function of the estimated blood oxygen saturation through the linear translation function, so that the fitting result of the mapping function of the estimated blood oxygen saturation matches the target user, and synchronizing the corrected mapping function of blood oxygen saturation to the blood oxygen calibration model.

2. The non-contact blood oxygen monitoring method based on mobile phone DC component calibration according to claim 1, characterized in that, The method of acquiring the photoplethysmography (PPG) signal of the target user's hand based on the constructed hand reflective optical path and extracting the DC component and hand pulsation component includes the following steps: The RGB video signal containing the hand reflection optical path constructed when the target user's hand covers the camera is acquired and processed by sliding window to obtain video data for each channel; A bandpass filter was used to extract the hand pulsation component of the video data of each channel, and a low-pass filter was used to remove the high-frequency noise of the video data of each channel. The mean value of the filtered data was used as the DC component of the hand in the corresponding time window.

3. The non-contact blood oxygen monitoring method based on mobile phone DC component calibration according to claim 2, characterized in that, The steps of acquiring the first facial video information of the target user, collecting the facial photoplethysmography (PPG) signal of the target user, and extracting the facial DC component include the following: First facial video information containing the target user's face is acquired and processed using a sliding window. A low-pass filter is used to extract the facial DC component within each time window.

4. The non-contact blood oxygen monitoring method based on mobile phone DC component calibration according to claim 3, characterized in that, The method of acquiring the hand photoplethysmography (PPG) signal of the target user based on the constructed hand reflective optical path and extracting the DC component and pulsation component of the hand; and acquiring the first facial video information of the target user, after acquiring the facial PPG signal of the target user and extracting the facial DC component, includes the following steps: The hand photoplethysmography (PPG) signals and facial PPG signals are aligned using timestamps to achieve time alignment of data and consistency control of physiological states.

5. The non-contact blood oxygen monitoring method based on mobile phone DC component calibration according to claim 1, characterized in that, After constructing the blood oxygen calibration model based on the hand DC component, hand pulsation component, and facial pulsation component, the process includes the following steps: creating a target user blood oxygen information database based on the first facial video information, mapping the blood oxygen calibration model to the target user blood oxygen information database, and calling the latest blood oxygen calibration model in the target user blood oxygen information database according to the correspondence between the second facial video information and the first facial video information.

6. The non-contact blood oxygen monitoring method based on mobile phone DC component calibration according to claim 1, characterized in that, The step of acquiring the first facial video information of the target user, collecting the facial photoplethysmography (PPG) signal of the target user and extracting the facial pulsation component, further includes the following steps: The face in the first facial video information is corrected, and the face region is divided into multiple regions of interest based on multiple facial muscle key points; Based on the pre-set region of interest, the corresponding key points of the facial muscles are selected, and the high reflectivity area for collecting facial photoplethysmography signals is selected. The facial photoplethysmography (PPG) signal of the high-reflectivity region is collected, and the facial pulsation component is extracted from it.

7. A non-contact blood oxygen monitoring system based on mobile phone DC component calibration, characterized in that, The system is used to execute the non-contact blood oxygen monitoring method based on mobile phone DC component calibration as described in any one of claims 1-6, and the system comprises: The acquisition module is used to acquire the photoplethysmography (PPG) signal of the target user's hand based on the constructed reflective optical path of the target user's hand; and to acquire the first facial video information of the target user, acquire the facial PPG signal of the target user and extract the facial pulsation component. A construction module is used to construct a blood oxygen calibration model based on the hand DC component, hand pulsation component, and facial pulsation component. The steps of constructing the blood oxygen calibration model based on the hand DC component, hand pulsation component, and facial pulsation component include: obtaining a standard blood oxygen saturation based on the relative amplitude ratio of the hand pulsation component and the hand DC component of each channel through a linear mapping relationship between the relative amplitude ratio and blood oxygen saturation; obtaining an estimated blood oxygen saturation based on the relative ratio of the facial DC component at different wavelengths through a linear mapping relationship between the relative ratio and blood oxygen saturation; calculating the deviation between the standard blood oxygen saturation and the estimated blood oxygen saturation to obtain a predicted deviation value; constructing a linear translation function based on the predicted deviation value; correcting the mapping function of the estimated blood oxygen saturation through the linear translation function so that the fitting result of the mapping function of the estimated blood oxygen saturation matches the target user; and synchronizing the corrected mapping function of blood oxygen saturation to the blood oxygen calibration model. The analysis module is used to acquire second facial video information, extract the DC component from the second facial video information and input it into the blood oxygen calibration model to obtain the current blood oxygen information of the target user.

8. A non-contact blood oxygen monitoring device based on mobile phone DC component calibration, comprising: At least one processor, and, A memory communicatively connected to the at least one processor; characterized in that the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the non-contact blood oxygen monitoring method based on mobile phone DC component calibration as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores, in the form of computer-readable instructions, a computer program implemented according to any one of claims 1 to 6, which, when invoked by a computer, executes the steps included in the corresponding method.

Images