Skin color self-adaptive wavelength adjusting method and device for fingertip pulse oximeter

Abstract

The invention discloses a skin color self-adaptive wavelength adjusting method and device for a fingertip pulse oximeter, and relates to the technical field of oximeters, the method comprises the following steps: when a finger of a user stretches into a detection cavity of the fingertip pulse oximeter, obtaining a finger image of the user through an image acquisition device arranged in the detection cavity; performing skin color analysis according to the user finger image to obtain a user finger skin color feature value; performing emitted light wavelength median optimization based on the user finger skin color characteristic value to obtain a red light band target wavelength and an infrared band target wavelength; and initializing the infrared transmitter according to the target wavelength of the red light band and the target wavelength of the infrared band. The technical problem that the blood oxygen detection precision is not matched with the skin color due to the fact that a traditional fingertip pulse oximeter cannot adjust the detection wavelength in real time according to the skin color of the user is solved, and the purposes that the blood oxygen content detection precision is improved by adaptively adjusting the target wavelengths of the red light band and the infrared band according to the finger skin color feature value of the user, and the user experience is improved are achieved. And requirements of users with different skin colors can be met.

Description

Technical Field

[0001] The present application relates to the technical field of blood oximeters, and in particular to a method and device for adjusting the skin color adaptive wavelength of a fingertip pulse oximeter. Background Art

[0002] Fingertip pulse oximeter is a device widely used in the field of medical and health monitoring. It evaluates blood oxygen saturation by measuring the oxygen content in the blood at the fingertips of the human body. In traditional fingertip pulse oximeters, fixed red light wavelengths (about 650nm) and infrared wavelengths (about 850nm) are usually used for detection. Although these fixed wavelengths can meet the needs of most users, there are obvious technical limitations in actual use. First, the skin color of different users varies significantly, and skin color directly affects the penetration depth and scattering state of light in the skin. For example, darker skin color has a greater impact on the absorption and scattering of light due to the presence of higher melanin content, thereby reducing the accuracy of the detection signal. This difference in skin color causes the accuracy deviation of the detection results of the oximeter in different populations. Secondly, traditional fingertip pulse oximeters are usually unable to dynamically adjust the wavelength to adapt to individual differences, and their preset wavelength design cannot cover the optimal optical response under a variety of skin color conditions. Therefore, in some special scenarios (such as users with dark skin or users with abnormal blood circulation), the detection accuracy and reliability may be significantly reduced, affecting its wide application in clinical and home scenarios. In addition, due to the use of fixed wavelengths, traditional oximeters may have difficulty ensuring the stability and accuracy of detection results in high-noise environments, especially among users with darker skin or complex finger shapes. This problem is more prominent. Summary of the invention

[0003] The present application provides a skin color adaptive wavelength adjustment method and device for a fingertip pulse oximeter, which solves the technical problem that a traditional fingertip pulse oximeter cannot adjust the detection wavelength in real time according to the user's skin color, resulting in the blood oxygen detection accuracy not being suitable for the skin color. The application achieves the technical effect of improving the blood oxygen content detection accuracy and adapting to the needs of users with different skin colors by adaptively adjusting the target wavelengths of the red light band and the infrared band according to the characteristic values ​​of the user's finger skin color.

[0004] The present application provides a skin color adaptive wavelength adjustment method for a fingertip pulse oximeter, the method comprising: when a user's finger is inserted into a detection cavity of the fingertip pulse oximeter, an image of the user's finger is obtained by an image acquisition device built into the detection cavity; skin color analysis is performed according to the image of the user's finger to obtain a skin color characteristic value of the user's finger; based on the skin color characteristic value of the user's finger, a median optimization of the wavelength of emitted light is performed to obtain a target wavelength of a red light band and a target wavelength of an infrared band; and an infrared emitter of the fingertip pulse oximeter is initialized according to the target wavelength of the red light band and the target wavelength of the infrared band.

[0005] The present application also provides a skin color adaptive wavelength adjustment device for a fingertip pulse oximeter, including: an image acquisition module, which is used to obtain a user's finger image through an image acquisition device built in the detection cavity when the user's finger extends into the detection cavity of the fingertip pulse oximeter; a skin color analysis module, which is used to perform skin color analysis based on the user's finger image to obtain the user's finger skin color characteristic value; a wavelength median optimization module, which is used to optimize the median value of the emission light wavelength based on the user's finger skin color characteristic value to obtain the target wavelength of the red light band and the target wavelength of the infrared band; an infrared emitter initialization module, which is used to initialize the infrared emitter of the fingertip pulse oximeter according to the target wavelength of the red light band and the target wavelength of the infrared band.

[0006] It is intended to solve the technical problem that the traditional fingertip pulse oximeter cannot adjust the detection wavelength in real time according to the user's skin color, resulting in the mismatch between the blood oxygen detection accuracy and the skin color, by means of the fingertip pulse oximeter skin color adaptive wavelength adjustment method and device proposed in the present application. When the user's finger extends into the detection cavity of the fingertip pulse oximeter, a user's finger image is obtained through an image acquisition device built in the detection cavity; skin color analysis is performed based on the user's finger image to obtain the user's finger skin color characteristic value; the median value of the emission light wavelength is optimized based on the user's finger skin color characteristic value to obtain the target wavelength of the red light band and the target wavelength of the infrared band; the infrared emitter of the fingertip pulse oximeter is initialized according to the target wavelength of the red light band and the target wavelength of the infrared band. The technical effect is achieved of adaptively adjusting the target wavelengths of the red light band and the infrared band according to the user's finger skin color characteristic value, improving the detection accuracy of blood oxygen content, and meeting the needs of users with different skin colors. Description of the Drawings

[0007] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings of the embodiments of the present disclosure will be briefly introduced below. Flowcharts are used in this application to illustrate the operations performed by the devices according to the embodiments of the present application. It should be understood that the operations in the front or below do not necessarily need to be executed precisely in sequence. On the contrary, according to the need, various steps can be executed in reverse order or simultaneously. At the same time, other operations can also be added to these processes, or one or several operations can be removed from these processes.

[0008] Figure 1 It is a schematic flowchart of the fingertip pulse oximeter skin color adaptive wavelength adjustment method provided by the embodiment of the present application.

[0009] Figure 2 It is a schematic structural diagram of the fingertip pulse oximeter skin color adaptive wavelength adjustment device provided by the embodiment of the present application.

[0010] Description of reference numerals: Image acquisition module 1, skin color analysis module 2, wavelength median optimization module 3, infrared emitter initialization module 4. Detailed implementation manners

[0011] The above description is only an overview of the technical solution of this application. In order to be able to understand the technical means of this application more clearly, it can be implemented according to the content of the description. And in order to make the above and other purposes, features and advantages of this application more obvious and understandable, the following specifically gives the detailed implementation manners of this application.

[0012] In order to make the purpose, technical solution and advantages of this application clearer, the following will further describe this application in detail with reference to the accompanying drawings. The described embodiments should not be regarded as limitations of this application. All other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of this application.

[0013] In the following description, "some embodiments" are involved, which describe a subset of all possible embodiments. However, it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict. The terms "first\second" involved are only used to distinguish similar objects and do not represent a specific order for the objects. The terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, device, product or server including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or modules not clearly listed or inherent to these processes, methods, products or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application.

[0014] The embodiment of this application provides a method for skin color adaptive wavelength adjustment of a fingertip pulse oximeter, as Figure 1 shown, the method includes: When the user's finger extends into the detection cavity of the fingertip pulse oximeter, a user finger image is obtained through an image acquisition device built in the detection cavity.

[0015] In the embodiments of the present application, when the user inserts a finger into the detection cavity of the fingertip pulse oximeter, the built-in image acquisition device of the device will be immediately activated to capture an image of the user's finger. The image acquisition device is usually a high-resolution micro camera or image sensor, which can clearly record the external features of the finger, including the color, shape and size of the finger. These image data will be transmitted to the processing unit of the device in real time for subsequent analysis, such as extracting the skin color feature value of the finger, etc., so as to provide basic data support for the subsequent optimization of the detection wavelength. This process is usually completed without the user's awareness, both quickly and accurately.

[0016] Perform skin color analysis based on the user finger image to obtain the user finger skin color feature value.

[0017] In one embodiment, through the acquired user finger image, the system terminal will analyze the image, focusing on identifying the skin color features of the finger. Specifically, the system terminal will first extract the color information in the image and process it through methods such as color aggregation and pixel statistics to remove background interference and focus on the finger area, and calculate the feature value representing the skin color of the user's finger. This skin color feature value is usually a quantified value, which can reflect the main skin color attributes of the user's finger and provide accurate data support for subsequent wavelength optimization. The entire analysis process combines efficiency and accuracy to ensure that users with different skin colors can obtain reliable detection results.

[0018] Furthermore, the present application provides skin color analysis based on the user finger image to obtain the user finger skin color feature value, including: Perform color neighborhood aggregation on the user finger image to obtain the first region, the second region until the Nth region; traverse the first region, the second region until the Nth region for finger shape matching to obtain the finger shape calibration region, where the finger shape calibration region belongs to the first region, the second region until the Nth region; extract the mode color feature value of the finger shape calibration region and set it as the user finger skin color feature value.

[0019] Preferably, after the user's finger image is captured, the system terminal converts the image from the RGB color space to the Lab color space because the Lab color space can better represent the color differences perceived by the human eye, which helps to improve the clustering effect. Subsequently, each pixel point in the image is converted into a color vector. For example, the three channel values (L, a, b) of the Lab color space are used as the feature vectors. In the color clustering stage, the K-means clustering algorithm is used to group the color vectors. To this end, the number of clustering categories K is set, which can be set empirically or dynamically calculated using methods such as the elbow method to determine the appropriate value of K. Then, K clustering centers are randomly initialized, and each center point represents a main color. For each pixel point, the Euclidean distance between it and all clustering centers is calculated, and it is assigned to the category of the nearest clustering center. After the initial assignment is completed, the mean value of the pixel points within each category is recalculated and used as the new clustering center. This process is repeated until the change amplitude of the clustering center is less than the set convergence threshold or the set number of iterations is reached. After clustering is completed, color neighborhood regions are formed according to the pixel grouping of each category, and a unique number is assigned to each neighborhood region to obtain the first region, the second region up to the Nth region. At the same time, the generated neighborhood regions are optimized to eliminate regions that are too small or isolated (such as regions with an area below the set threshold), thereby reducing the interference of invalid regions. Then, all color neighborhood regions are traversed, and finger shape matching analysis is performed one by one to identify the region where the user's finger is located. In this process, the system terminal constructs a finger shape mold according to the user input and preset parameters, and performs shape matching between the mold and each color neighborhood region to calculate their similarity degree, obtaining the similarity value of each region. Then, the finger region is screened according to the similarity value threshold to obtain the finger shape calibration region. This finger shape calibration region can be any one of the first region, the second region up to the Nth region. In the determined finger shape calibration region, the system terminal performs statistical analysis on all pixel color values in the region, and extracts the main color feature of the region by calculating the mode of the color values (i.e., the color value that appears most frequently), ensuring that this feature value can accurately represent the skin color of the user's finger. The system terminal uses this mode color feature value as the final skin color feature value of the user's finger, providing basic data for subsequent wavelength optimization. Through the above steps, the skin color feature value of the user's finger can be accurately extracted, ensuring that the wavelength adjustment adapts to individual skin color differences and improving the reliability and accuracy of detection.

[0020] Further, the present application provides traversing the first region, the second region up to the Nth region for finger shape matching to obtain the finger shape calibration region, including: Receive the user finger width input by the user terminal, and obtain the finger accommodating length of the detection cavity; based on the user finger width and the finger accommodating length, construct a finger mold; traverse the first region, the second region until the Nth region, and perform shape similarity analysis with the finger mold respectively to obtain the first shape similarity value, the second shape similarity value until the Nth shape similarity value; extract the regions in the first shape similarity value, the second shape similarity value until the Nth shape similarity value that are greater than or equal to the shape similarity threshold value, and set them as candidate calibration regions; when the number of candidate calibration regions is 1, set the candidate calibration region as the finger calibration region; when the number of candidate calibration regions is not 1, update the user finger image and execute the skin color analysis loop.

[0021] Optionally, when the user inserts a finger into the detection cavity, the system terminal will receive the width of the user's finger shape input by the user end, usually provided by measurement or directly input by the user. At the same time, the finger shape tolerable length of the detection cavity will be automatically obtained according to internal parameters or sensors. These two data will serve as the basis for constructing a finger shape mold. Using the width of the user's finger shape and the finger shape tolerable length of the detection cavity, a finger shape mold is constructed through 3D modeling tools. Subsequently, the color neighborhood regions (the first region, the second region up to the Nth region) in the user finger image are traversed. For each color neighborhood region, the system terminal compares it with the constructed finger shape mold one by one and performs shape similarity analysis. Specifically, the system terminal first extracts the boundary of each color neighborhood region to obtain a set of boundary points of the region, and these point sets completely describe the external contour of the region. At the same time, the set of boundary points of the finger shape mold has been generated through 3D modeling tools to simulate the natural contour of the finger. For each color neighborhood region, a shape similarity analysis is performed between its set of boundary points and the set of boundary points of the finger shape mold. Here, the Hausdorff distance is used to quantify the degree of shape difference between the region boundary and the mold boundary. Other distances such as the Frechet distance and the Euclidean distance can also be used. The Hausdorff distance is a commonly used distance metric method for calculating the closest distance of the farthest matching points between two point sets, which can effectively describe the maximum deviation between the region and the mold. By calculating the Hausdorff distance between each color neighborhood region and the finger shape mold, and then using 1 minus the ratio of this distance to the maximum distance, a similarity value is obtained as the shape similarity value of this region. This shape similarity value ranges from 0 to 1, where 1 indicates that the shapes are exactly the same, and the closer the value is to 1, the closer the shape of the region is to the finger shape mold. After that, the system terminal traverses all color neighborhood regions, calculates the similarity values one by one, records the results of each region, and thus obtains the first shape similarity value, the second shape similarity value, up to the Nth shape similarity value. These similarity values are compared with a preset shape similarity value threshold, and the regions greater than or equal to the shape similarity value threshold are selected and marked as candidate calibration regions. If the number of candidate calibration regions is 1, then this region is directly set as the finger shape calibration region. If the number of candidate calibration regions is more than 1, it means that there are multiple possible matching regions. At this time, the system terminal will enter the further analysis stage, update the user finger image (such as by enhancing the contrast or refining the edge features) and re-execute the skin color analysis loop. This loop will be repeated until a unique finger shape calibration region is selected to ensure the accuracy of the target region. Through the above process, the specific region where the user's finger is located can be efficiently identified, providing accurate region positioning for subsequent skin color analysis and wavelength optimization.

[0022] Based on the user finger skin color feature value, the median value of the emitted light wavelength is optimized to obtain the target wavelength in the red light band and the target wavelength in the infrared band.

[0023] In one embodiment, the red light band is generally defined in the range of 600 - 750 nm, and the conventional preset value is 650 nm, which is the band where deoxyhemoglobin has the highest light absorption efficiency. The infrared band is generally in the range of 850 - 1000 nm, and the conventional preset value is 850 nm, which is the band where oxyhemoglobin has the highest light absorption efficiency. Although these preset values can provide good detection effects under standard conditions, due to the differences in user skin colors, the penetration depth and scattering state of light in skin tissue will be significantly affected, thus affecting the detection accuracy. To solve this problem, the system terminal will perform median optimization on the light wavelengths of the above bands based on the skin color characteristic values of the user's finger. In this process, the median reconstruction of the red light wavelength limit interval and the infrared light wavelength limit interval will be performed according to the skin color characteristic values of the user's finger to obtain the red light wavelength reconstruction limit interval and the infrared light wavelength reconstruction limit interval. For example, for the red light band (600 - 750 nm), if the skin color characteristic value indicates that the best penetration range should be biased towards higher wavelengths, the limit interval may be reconstructed as 660 - 700 nm. After obtaining the red light wavelength reconstruction limit interval and the infrared light wavelength reconstruction limit interval, the system terminal will find the optimal median wavelength from the red light wavelength reconstruction limit interval and the infrared light wavelength reconstruction limit interval as the target wavelength of the red light band and the target wavelength of the infrared light band. By this method, the emission wavelengths of red light and infrared light can be dynamically adjusted to better adapt to the skin color characteristics of the user, thereby maximizing the light absorption efficiency and improving the accuracy and reliability of blood oxygen detection.

[0024] Furthermore, the present application provides median optimization of the emission light wavelength based on the skin color characteristic values of the user's finger to obtain the target wavelength of the red light band and the target wavelength of the infrared light band, including: Obtaining the red light wavelength limit interval and the infrared light wavelength limit interval; performing median reconstruction on the red light wavelength limit interval and the infrared light wavelength limit interval based on the skin color characteristic values of the user's finger to obtain the red light wavelength reconstruction limit interval and the infrared light wavelength reconstruction limit interval; extracting the target wavelength of the red light band with the smallest deviation from 650 nm from the red light wavelength reconstruction limit interval; extracting the target wavelength of the infrared light band with the smallest deviation from 850 nm from the infrared light wavelength reconstruction limit interval.

[0025] Preferably, the system terminal defines an initial wavelength range according to the standard light absorption characteristics. The red light wavelength limit interval can be 600 - 750 nm, which covers the region with the highest absorbance of deoxyhemoglobin. The infrared light wavelength limit interval can be 850 - 1000 nm, which covers the region with the highest absorbance of oxyhemoglobin. Subsequently, according to the skin color characteristic value of the user's finger, the system terminal will perform median reconstruction on the limit intervals of the red and infrared wavelengths based on the penetration ratio calculated by dividing the first penetration depth by the user's finger thickness. This is because darker skin may cause a reduction in the penetration depth of light, while lighter skin reflects more light. Therefore, it is necessary to reconstruct the limit intervals to improve the penetration efficiency. After median reconstruction, a new red light wavelength reconstruction limit interval and an infrared light wavelength reconstruction limit interval are obtained respectively. Within the red light wavelength reconstruction limit interval, the system terminal searches for the wavelength with the smallest deviation from the standard preset value of 650 nm as the target wavelength in the red light band by means of the absolute difference. Similarly, within the infrared light wavelength reconstruction limit interval, the wavelength with the smallest deviation from the standard preset value of 850 nm is searched for as the target wavelength in the infrared light band. After completing the above steps, the system terminal uses the determined target wavelength in the red light band and the target wavelength in the infrared light band as the final output to initialize the light emission module of the device. Through dynamic optimization, the emitted light wavelength is more suitable for the skin color characteristics of the user, so as to improve the accuracy and reliability of detection.

[0026] Furthermore, the present application provides median reconstruction of the red light wavelength limit interval and the infrared light wavelength limit interval based on the skin color characteristic value of the user's finger to obtain a red light wavelength reconstruction limit interval and an infrared light wavelength reconstruction limit interval, including: Extract the lower limit wavelength of the red light wavelength limit interval, extract the upper limit wavelength of the infrared light wavelength limit interval, and construct a first median optimization interval; take the median of the first median optimization interval to obtain a first median wavelength; predict the penetration depth according to the skin color characteristic value of the user's finger and the first median wavelength to obtain a first penetration depth; receive the user's finger thickness from the user terminal; calculate the ratio of the first penetration depth to the user's finger thickness, denoted as the penetration ratio; when the penetration ratio belongs to the penetration ratio threshold interval, set the first median wavelength as the lower limit wavelength threshold, and use the lower limit wavelength threshold as the lower limit wavelength to reconstruct the red light wavelength limit interval and the infrared light wavelength limit interval to obtain the red light wavelength reconstruction limit interval and the infrared light wavelength reconstruction limit interval.

[0027] Optionally, the system terminal first extracts the lower wavelength limit (e.g., 600 nm) from the red light wavelength limit range and extracts the upper wavelength limit (e.g., 1000 nm) from the infrared wavelength limit range. These two values will be used as the basis for subsequent optimization calculations. Then, using the extracted lower red light wavelength (e.g., 600 nm) and upper infrared wavelength (e.g., 1000 nm) as boundaries, a unified range containing the information of the two wavelength bands is formed, namely the first median optimization range. Subsequently, a median operation is performed on the first median optimization range to calculate the first median wavelength. For example, for the range [600 nm, 1000 nm], the first median wavelength is (600 + 1000) / 2 = 800 nm. Subsequently, combining the user's finger skin color characteristic value and the first median wavelength, prediction is performed through a pre-trained penetration depth prediction model. This model is trained based on experimental data. The model processes the received user's finger skin color characteristic value and the first median wavelength and outputs the first penetration depth, indicating the depth at which light penetrates the user's finger at the first median wavelength. After that, the finger thickness data input by the user terminal is received, usually directly obtained by a measuring device or provided by the user. For example, the finger thickness may be 8 mm. The system terminal calculates the ratio of the first penetration depth to the user's finger thickness and uses it as the penetration ratio. This ratio reflects whether the light can effectively penetrate the user's finger at the current wavelength. For example, if the first penetration depth is 6 mm and the user's finger thickness is 8 mm, the penetration ratio is 6 / 8 = 0.75. After calculating the penetration ratio, the system terminal checks whether the penetration ratio falls within a preset penetration ratio threshold range (e.g., 0.7 - 0.9). If the penetration ratio is not within the threshold range, the system terminal adjusts the limit range and recalculates, which may specifically include narrowing the range or adjusting the median. If the penetration ratio is within the threshold range, the first median wavelength is used as the lower wavelength threshold of the red light wavelength limit range. At the same time, a similar adjustment is made to the infrared wavelength limit range to improve the penetration efficiency. Through the above steps, the optimized red light wavelength reconstruction limit range and infrared wavelength reconstruction limit range are finally obtained. These new limit ranges can better adapt to the skin color characteristics and thickness conditions of the user's finger and provide support for the extraction of subsequent wavelength targets.

[0028] Furthermore, the present application provides predicting the penetration depth according to the user's finger skin color characteristic value and the first median wavelength to obtain the first penetration depth, including: Collecting finger red light penetration experimental data, where the finger red light penetration experimental data includes a skin color characteristic vector variable data set, a penetration red light wavelength variable data set, and a penetration depth calibration data set; using the penetration depth calibration data set as output supervision, configuring a random forest with the skin color characteristic vector variable data set and the penetration red light wavelength variable data set as inputs to obtain a penetration depth prediction model; and processing the user's finger skin color characteristic value and the first median wavelength according to the penetration depth prediction model to obtain the first penetration depth.

[0029] Optionally, the system terminal establishes a penetration depth prediction model by collecting finger red light penetration experiment data, where the experiment data includes a skin color feature vector variable data set, a penetrated red light wavelength variable data set, and a penetration depth calibration data set. The experiment data is collected through an experimental device to ensure coverage of a variety of skin colors and wavelength conditions, and the reliability of the data is guaranteed by the reflectance, absorptance, and penetration depth calibration values measured by a spectral analyzer and a photodetector. After collection, the data is cleaned and formatted, outliers are removed, missing data is filled, and normalization is performed. The skin color feature vector and the red light wavelength are used as input variables, and the penetration depth calibration value is used as the output target. Subsequently, the data set is divided into a training set, a validation set, and a test set to ensure that each subset covers different skin color features and wavelength ranges. During the model training process, random forest is selected as the core algorithm of the penetration depth prediction model, and a deep neural network, etc. can also be selected. Before training, model parameters are set, including the number of decision trees, the maximum depth of each tree, and the random sampling ratio. At the start of training, samples are randomly drawn with replacement from the training set to generate multiple sub-data sets, and each sub-data set is used to train a decision tree. For each decision tree, the feature space is gradually split according to the skin color feature vector and the red light wavelength variable, and the feature and split point with the highest node purity after splitting are selected. The node purity uses the mean squared error as an index to ensure the optimal fitting effect of the predicted continuous variable. This splitting process continues until the maximum depth is reached or the number of node samples is less than a preset threshold. After that, the prediction results of all decision trees are fused by the average value to form the output of the final random forest model. During the training process, the hyperparameters of the model are optimized through the validation set, including adjusting the number of decision trees, the sample and feature sampling ratios, and the maximum tree depth. The optimization process can use grid search or random search methods to select the optimal parameter combination, and at the same time calculate the mean squared error (MSE) or root mean squared error (RMSE) to verify the model performance. After model validation, the test set is used to evaluate the prediction effect of the final model, and the mean absolute error (MAE), mean squared error (MSE), and coefficient of determination ( ), etc. are used as performance indicators to ensure the generalization ability of the model under different skin color and wavelength conditions. Finally, based on the constructed random forest model, the actual data of the user's finger is predicted. The skin color feature value and the first median wavelength of the user's finger are used as inputs, and the model processes the input data using the learned relationship between skin color and wavelength on the penetration depth, and outputs the first penetration depth. This result reflects the effective penetration depth of light under the current skin color feature and wavelength conditions, providing accurate data support for subsequent wavelength optimization, thereby ensuring the accuracy and adaptability of blood oxygen detection.

[0030] Furthermore, the method further includes: When the penetration ratio is greater than the upper limit value of the penetration ratio threshold range, a second median optimization range is constructed based on the lower wavelength of the red light wavelength limit range in combination with the first median wavelength; median reconstruction loop analysis is performed according to the second median optimization range.

[0031] Optionally, when the penetration ratio is greater than the upper limit value of the preset penetration ratio threshold range (for example, the threshold range is 0.7 - 0.9 and the current penetration ratio is 0.95), it indicates that the penetration depth of the current light wavelength is too high relative to the finger thickness, which may affect the detection accuracy. At this time, it is necessary to re-optimize the red light wavelength limit range and the first median wavelength to determine a more suitable wavelength range. First, the lower wavelength of the red light wavelength limit range and the first median wavelength are extracted, and a new optimization range, that is, the second median optimization range, is constructed with these two values as boundaries. This range represents a red light wavelength range that may be more suitable for the user's skin color characteristics and finger thickness. Subsequently, the aforementioned median reconstruction analysis is performed again within this second median optimization range. This analysis process will loop continuously, and the system terminal calculates a new median wavelength and evaluates its penetration depth each time, gradually converging the optimization range until the penetration ratio falls within the preset threshold range. The finally determined median wavelength will be used as a key parameter for the reconstructed limit range to optimize the red light wavelength range to make it more adaptable to the characteristics of the current user's finger and the detection requirements.

[0032] Initialize the infrared emitter of the fingertip pulse oximeter according to the target wavelength of the red light band and the target wavelength of the infrared light band.

[0033] In one embodiment, according to the determined target wavelength of the red light band and the target wavelength of the infrared light band, the system terminal initializes the infrared emitter of the fingertip pulse oximeter. The infrared emitter in the fingertip pulse oximeter is usually composed of a tunable light source or a filter component, which can dynamically adjust the wavelength of the emitted light within a certain range. The system terminal first inputs the target wavelength of the red light band and the target wavelength of the infrared light band as initialization parameters into the control module respectively. Subsequently, the control module adjusts the wavelength output of the infrared emitter according to the input target wavelength. For example, by changing the excitation frequency of the emission light source or adjusting the settings of the wavelength selector (such as a liquid crystal tunable filter or an acousto-optic tunable filter), the emitter outputs precise red light and infrared light wavelengths. After initialization, the emitter can stably output red light and infrared light consistent with the target wavelengths, providing the required light source for subsequent blood oxygen detection. This process ensures the adaptive adjustment of the device to the user's skin color and finger characteristics, which helps to improve the detection accuracy and applicability.

[0034] In the foregoing, reference is made to Figure 1 The method for adaptively adjusting the wavelength according to the skin color of the fingertip pulse oximeter according to the embodiments of the present invention is described in detail. Next, reference will be made to Figure 2Describe a skin color adaptive wavelength adjustment device for a fingertip pulse oximeter according to an embodiment of the present invention.

[0035] The skin color adaptive wavelength adjustment device for a fingertip pulse oximeter according to an embodiment of the present invention is used to solve the technical problem that traditional fingertip pulse oximeters cannot adjust the detection wavelength in real time according to the user's skin color, resulting in the mismatch between the blood oxygen detection accuracy and the skin color. It achieves the technical effect of adaptively adjusting the target wavelengths of the red light band and the infrared band according to the skin color characteristic values of the user's finger, improving the detection accuracy of blood oxygen content, and meeting the needs of users with different skin colors. The skin color adaptive wavelength adjustment device for a fingertip pulse oximeter includes: an image acquisition module 1, a skin color analysis module 2, a wavelength median optimization module 3, and an infrared emitter initialization module 4.

[0036] The image acquisition module 1 is configured to obtain an image of the user's finger through an image acquisition device built in the detection cavity when the user's finger is inserted into the detection cavity of the fingertip pulse oximeter; the skin color analysis module 2 is configured to perform skin color analysis based on the image of the user's finger to obtain the skin color characteristic values of the user's finger; the wavelength median optimization module 3 is configured to optimize the median wavelength of the emitted light based on the skin color characteristic values of the user's finger to obtain the target wavelengths of the red light band and the infrared band; the infrared emitter initialization module 4 is configured to initialize the infrared emitter of the fingertip pulse oximeter according to the target wavelengths of the red light band and the infrared band.

[0037] Further, the skin color analysis module 2 further includes: Perform color neighborhood aggregation on the image of the user's finger to obtain the first region, the second region, up to the Nth region; traverse the first region, the second region, up to the Nth region for finger shape matching to obtain a finger shape calibration region, where the finger shape calibration region belongs to the first region, the second region, up to the Nth region; extract the mode color characteristic value of the finger shape calibration region and set it as the skin color characteristic value of the user's finger.

[0038] Further, the skin color analysis module 2 further includes: Receive the user's finger width input by the user terminal, and obtain the finger accommodating length of the detection cavity; based on the user's finger width and the finger accommodating length, construct a finger mold; traverse the first region, the second region until the Nth region, and perform shape similarity analysis with the finger mold respectively to obtain the first shape similarity value, the second shape similarity value until the Nth shape similarity value; extract the regions in the first shape similarity value, the second shape similarity value until the Nth shape similarity value that are greater than or equal to the shape similarity threshold value, and set them as the candidate calibration regions; when the number of candidate calibration regions is 1, set the candidate calibration region as the finger calibration region; when the number of candidate calibration regions is not 1, update the user finger image and execute the skin color analysis loop.

[0039] Further, the wavelength median optimization module 3 further includes: Obtain the red light wavelength limit interval and the infrared light wavelength limit interval; perform median reconstruction on the red light wavelength limit interval and the infrared light wavelength limit interval based on the user's finger skin color characteristic value to obtain the red light wavelength reconstructed limit interval and the infrared light wavelength reconstructed limit interval; extract the red light band target wavelength with the smallest deviation from 650nm from the red light wavelength reconstructed limit interval; extract the infrared light band target wavelength with the smallest deviation from 850nm from the infrared light wavelength reconstructed limit interval.

[0040] Further, the wavelength median optimization module 3 further includes: Extract the lower limit wavelength of the red light wavelength limit interval, extract the upper limit wavelength of the infrared light wavelength limit interval, and construct a first median optimization interval; take the median of the first median optimization interval to obtain a first median wavelength; perform penetration depth prediction based on the user's finger skin color characteristic value and the first median wavelength to obtain a first penetration depth; receive the user's finger thickness input by the user terminal; calculate the ratio of the first penetration depth to the user's finger thickness, and set it as the penetration ratio; when the penetration ratio belongs to the penetration ratio threshold interval, set the first median wavelength as the lower limit wavelength threshold, and use the lower limit wavelength threshold as the lower limit wavelength to reconstruct the red light wavelength limit interval and the infrared light wavelength limit interval to obtain the red light wavelength reconstructed limit interval and the infrared light wavelength reconstructed limit interval.

[0041] Further, the wavelength median optimization module 3 further includes: When the penetration ratio is greater than the upper limit value of the penetration ratio threshold interval, based on the lower limit wavelength of the red light wavelength limit interval, combine it with the first median wavelength to construct a second median optimization interval; perform median reconstruction loop analysis according to the second median optimization interval.

[0042] Further, the wavelength median optimization module 3 further includes: Collect the experimental data of red light penetration through the finger. Among them, the experimental data of red light penetration through the finger includes a skin color feature vector variable data set, a penetrated red light wavelength variable data set, and a penetration depth calibration data set. Using the penetration depth calibration data set as output supervision, configure a random forest with the skin color feature vector variable data set and the penetrated red light wavelength variable data set as inputs to obtain a penetration depth prediction model. According to the penetration depth prediction model, process the skin color feature value of the user's finger and the first median wavelength to obtain the first penetration depth.

[0043] The skin color adaptive wavelength adjustment device of the fingertip pulse oximeter provided by the embodiments of the present invention can execute the skin color adaptive wavelength adjustment method of the fingertip pulse oximeter provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0044] Although the present application makes various references to certain modules in the device according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or the server. The included individual units and modules are only divided according to functional logic, but are not limited to the above division as long as the corresponding functions can be achieved. In addition, the specific names of the functional units are only for easy distinction from each other and do not limit the protection scope of the present invention.

[0045] The above specific implementation manners do not constitute a limitation to the protection scope of the present application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of the present application shall be included within the protection scope of the present application.

Claims

1. A method for adjusting the skin color adaptive wavelength of a fingertip pulse oximeter, characterized in that: include: When a user's finger is inserted into the detection cavity of the fingertip pulse oximeter, an image of the user's finger is obtained through an image acquisition device built into the detection cavity; Perform skin color analysis based on the user's finger image to obtain a skin color feature value of the user's finger; Based on the skin color feature value of the user's finger, the median wavelength of the emitted light is optimized to obtain a target wavelength in the red light band and a target wavelength in the infrared band; The infrared emitter of the fingertip pulse oximeter is initialized according to the red light band target wavelength and the infrared band target wavelength.

2. The method according to claim 1, characterized in that Performing skin color analysis according to the user's finger image to obtain a skin color feature value of the user's finger includes: Performing color neighborhood aggregation on the user's finger image to obtain a first region, a second region, and finally an Nth region; Traversing the first area, the second area, and up to the Nth area to perform finger shape matching, and obtaining a finger shape calibration area, wherein the finger shape calibration area belongs to the first area, the second area, and up to the Nth area; The mode color feature value of the finger shape calibration area is extracted and set as the skin color feature value of the user's finger.

3. The method according to claim 2, characterized in that Traversing the first area, the second area, and finally the Nth area to perform finger shape matching to obtain a finger shape calibration area, including: Receiving a user's finger width input by a user terminal, and obtaining a finger-accommodating length of the detection cavity; constructing a finger mold based on the user's finger width and the finger's tolerable length; Traversing the first area, the second area, and up to the Nth area, respectively performing shape analysis with the finger-shaped mold to obtain a first shape value, a second shape value, and up to the Nth shape value; Extracting an area among the first shape-like value, the second shape-like value, and the Nth shape-like value that is greater than or equal to the shape-like value threshold, and setting it as a to-be-selected calibration area; When the number of the calibration areas to be selected is 1, the calibration areas to be selected are set as the finger-shaped calibration areas; When the number of the to-be-selected calibration areas is not 1, the user's finger image is updated to execute a skin color analysis loop.

4. The method according to claim 1, characterized in that The median wavelength of the emitted light is optimized based on the skin color characteristic value of the user's finger to obtain a target wavelength in the red light band and a target wavelength in the infrared band, including: Obtaining a red light wavelength limit interval and an infrared wavelength limit interval; Perform median reconstruction on the red light wavelength limit interval and the infrared wavelength limit interval based on the skin color feature value of the user's finger to obtain a red light wavelength reconstruction limit interval and an infrared wavelength reconstruction limit interval; Extracting the red light band target wavelength with the smallest deviation from 650nm from the red light wavelength reconstruction limit interval; The infrared band target wavelength with the smallest deviation from 850 nm is extracted from the infrared wavelength reconstruction restriction interval.

5. The method according to claim 4, characterized in that The red light wavelength limit interval and the infrared wavelength limit interval are median reconstructed based on the skin color feature value of the user's finger to obtain a red light wavelength reconstruction limit interval and an infrared wavelength reconstruction limit interval, including: Extracting the lower limit wavelength of the red light wavelength limit interval, extracting the upper limit wavelength of the infrared wavelength limit interval, and constructing a first median optimization interval; Taking the median of the first median optimization interval to obtain a first median wavelength; Predicting the penetration depth according to the skin color feature value of the user's finger and the first median wavelength to obtain a first penetration depth; receiving the user's finger thickness at the user end; Calculate the ratio of the first penetration depth to the thickness of the user's finger shape, and set it as the penetration ratio; When the transmittance ratio belongs to the transmittance ratio threshold interval, the first median wavelength is set as the lower limit wavelength threshold, and the lower limit wavelength threshold is used as the lower limit wavelength, and the red light wavelength limit interval and the infrared wavelength limit interval are reconstructed to obtain the red light wavelength reconstruction limit interval and the infrared wavelength reconstruction limit interval.

6. The method according to claim 5, characterized in that Also includes: When the transmittance is greater than the upper limit of the transmittance threshold interval, a second median optimization interval is constructed based on the lower limit wavelength of the red light wavelength limit interval and the first median wavelength; A median reconstruction loop analysis is performed according to the second median optimization interval.

7. The method according to claim 5, characterized in that Predicting the penetration depth according to the skin color feature value of the user's finger and the first median wavelength to obtain a first penetration depth includes: Collecting finger red light penetration experimental data, wherein the finger red light penetration experimental data includes a skin color feature vector variable data set, a penetration red light wavelength variable data set, and a penetration depth calibration data set; The penetration depth calibration data set is used as output supervision, and the skin color feature vector variable data set and the penetration red light wavelength variable data set are used as input to perform random forest configuration to obtain a penetration depth prediction model; According to the penetration depth prediction model, the user's finger skin color feature value and the first median wavelength are processed to obtain the first penetration depth.

8. A method for adjusting the skin color adaptive wavelength of a fingertip pulse oximeter, characterized in that: The device is used to implement the fingertip pulse oximeter skin color adaptive wavelength adjustment method according to any one of claims 1 to 7, comprising: Image acquisition module: when the user's finger is inserted into the detection cavity of the fingertip pulse oximeter, the image of the user's finger is obtained through the image acquisition device built into the detection cavity; Skin color analysis module: performs skin color analysis based on the user's finger image to obtain a skin color feature value of the user's finger; Wavelength median optimization module: performs wavelength median optimization of the emitted light based on the skin color feature value of the user's finger to obtain the target wavelength of the red light band and the target wavelength of the infrared band; Infrared transmitter initialization module: initializes the infrared transmitter of the fingertip pulse oximeter according to the red light band target wavelength and the infrared band target wavelength.

Images