Method and device for measuring oxygen saturation in blood

Abstract

The present disclosure generally relates to a computerized method (100) and device for measuring the oxygen saturation in a user's blood. A PPG signal is measured from light of one or more different wavelengths (110), and each PPG signal has a pulsatile component and a non-pulsatile component. For each PPG signal, a gradient of the non-pulsatile component of the PPG signal with respect to the light intensity of the wavelength is calculated (120). The user's skin tone (skin color) is determined from one or more gradients of the PPG signal and a first data model (130). A modulation ratio is calculated from the pulsatile and non-pulsatile components of a pair of PPG signals measured from different wavelengths (140), and a second data model is selected based on the user's skin tone (150). The oxygen saturation in the user's blood is determined from the modulation ratio and the selected second data model (160).

Description

Technical Field

[0001] This disclosure claims the benefit of Singapore Patent Application No. 10202250046F filed on May 30, 2022, which is incorporated herein by reference in its entirety.

[0002] This disclosure generally relates to the measurement of oxygen saturation in blood. More specifically, this disclosure describes various embodiments of methods and devices for measuring the oxygen saturation in a user's blood.

Background Art

[0003] The oxygen saturation in blood, i.e., the SpO2 level, can be used to detect various health conditions or disorders. For example, Obstructive Sleep Apnea (OSA) is a sleep-related breathing disorder that typically occurs when part or all of the upper airway is blocked during sleep. This leads to a decrease in SpO2 levels, which can drop by more than 40% in severe cases. Studies have also found correlations between OSA and other conditions and diseases such as cardiovascular disease (CVD), diabetes, and mental stress. SpO2 levels can also be used as an important indicator for respiratory diseases such as COVID-19.

[0004] SpO2 levels are generally measured using a pulse oximeter, which tends to overestimate SpO2 levels for people with dark skin. The pulse oximeter may indicate a normal SpO2 level, but the true SpO2 level may be lower. As a result, the user may not be aware that they are suffering from a low SpO2 level, which can be dangerous as there is a higher risk of hypoxemia, i.e., low blood oxygen.

[0005] Recent studies have shown that patients with darker skin have approximately three times the frequency of latent hypoxemia that goes undetected by pulse oximetry compared to patients with lighter skin. Latent hypoxemia is a condition where the arterial oxygen saturation is less than 88% despite the oxygen saturation on pulse oximetry being between 92% and 96%. Since pulse oximetry is widely used in medical decision-making, relying on pulse oximetry to triage patients and adjust supplemental oxygen levels can put patients with darker skin at increased risk of hypoxemia.

[0006] Accordingly, it is desirable to provide an improved method and device for measuring oxygen saturation in the blood to address or mitigate at least one of the above-mentioned problems and / or drawbacks. SUMMARY OF THE INVENTION

[0007] According to a first aspect of the present disclosure, there is a computerized method and a measuring device for measuring the oxygen saturation in a user's blood. The method includes measuring a set of photoplethysmography (PPG) signals from light of one or more different wavelengths, each PPG signal being measured from a respective wavelength and having a pulsatile component and a non-pulsatile component, calculating, for each PPG signal, a gradient of the non-pulsatile component of the PPG signal with respect to the light intensity of the respective wavelength, determining the user's skin tone from one or more gradients of the set of PPG signals and a first data model, calculating a modulation ratio from the pulsatile and non-pulsatile components of a pair of PPG signals measured from a pair of different wavelengths of light, selecting one from a plurality of second data models based on the user's skin tone, and determining the oxygen saturation in the user's blood from the modulation ratio and the selected second data model for the user's skin tone.

[0008] According to a first aspect of the present disclosure, there is a computerized method and a measuring device for determining a user's skin tone. The method includes measuring a set of photoplethysmography (PPG) signals from light of one or more different wavelengths, each PPG signal being measured from a respective wavelength and having a pulsatile component and a non-pulsatile component, and for each PPG signal, calculating a gradient of the non-pulsatile component of the PPG signal with respect to the light intensity of the respective wavelength, and determining the user's skin tone from one or more gradients of the set of PPG signals and a first data model.

[0009] A method and device for determining skin tone in accordance with the present disclosure and measuring oxygen saturation in the blood are thus disclosed herein. Various features, aspects, and advantages of the present disclosure will become even more apparent from the following detailed description of embodiments of the present disclosure, taken in conjunction with the accompanying drawings, which are presented by way of non-limiting example only.

Brief Description of the Drawings

[0010]

Figure 1

Figure 2A

Figure 2B

Figure 2C

Figure 2D

Figure 2E

Figure 3A

Figure 3B

Figure 3C

Figure 3D

Figure 3E

Figure 3F

Figure 3G

Figure 3H

Figure 3I

Figure 3J

Figure 3K

Figure 3L

Figure 4A

Figure 4B

Figure 5A

Figure 5B

Figure 6A

Figure 6B

Figure 7A

Figure 7B

Figure 8A

Figure 8B

Figure 8C

Figure 8D

Figure 8E

Figure 9

Figure 10

Figure 11A

Figure 11B

Figure 11C

Figure 11D

DETAILED DESCRIPTION OF THE INVENTION

[0011] For the purposes of brevity and clarity, the description of the embodiments of the present disclosure is directed, with reference to the drawings, to methods and devices for determining skin tone and measuring oxygen saturation in the blood. Aspects of the present disclosure are described in conjunction with the embodiments provided herein, but it is to be understood that they are not intended to limit the present disclosure to these embodiments. Rather, the present disclosure is intended to cover alternatives, modifications, and equivalents to the embodiments described herein, which are within the scope of the present disclosure as defined by the appended claims. Further, in the following detailed description, specific details are set forth in order to provide a thorough understanding of the present disclosure. However, it will be recognized by those of ordinary skill in the art, i.e., those skilled in the art, that the present disclosure can be practiced without these specific details and / or can be practiced with a plurality of details resulting from combinations of aspects of specific embodiments. In many cases, well-known systems, methods, procedures, and components are not described in detail so as not to unnecessarily obscure aspects of the embodiments of the present disclosure.

[0012] In embodiments of the present disclosure, the depiction of a given element in a particular figure or the consideration or use of a particular element number, or reference thereto in a corresponding description, may encompass the same, equivalent, or similar element or element number identified in another figure or the associated description.

[0013] References to "one embodiment / example", "another embodiment / example", "some embodiments / examples", "some other embodiments / examples", etc., indicate that the (one or more) embodiments / (one or more) examples so described may include certain features, structures, characteristics, properties, elements, or limitations, but not all embodiments / examples necessarily include that particular feature, structure, characteristic, property, element, or limitation. Also, the repeated use of the phrase "in one embodiment / example" or "in another embodiment / example" does not necessarily refer to the same embodiment / example.

[0014] The terms "have", "include", "possess", and the like do not preclude the presence of features / elements / steps other than those recited in the embodiments. The description of particular features / elements / steps in different embodiments does not indicate that combinations of those features / elements / steps may not be used in one embodiment.

[0015] As used herein, the terms "a" and "an" are defined as one or more than one. The use of " / " in a figure or related text is understood to mean "and / or" unless otherwise noted. The term "set" is defined as a non-empty finite compilation of elements that mathematically indicates at least one cardinality according to known mathematical definitions (e.g., the sets defined herein can correspond to units, singletons, or single-element sets, or multi-element sets). The recitation of a particular numerical value or numerical range herein is understood to include or describe approximate numerical values or numerical ranges. The terms "first", "second", etc. are used merely as labels or identifiers and are not intended to impose numerical requirements on their associated terms.

[0016] In representative or exemplary embodiments of the present disclosure, as shown in FIG. 1, there is a computer-implemented or computerized method 100 for measuring the oxygen saturation in a user's blood. The method can be executed on a measurement device having a processor, and various steps of the computerized method are executed in response to non-transitory instructions operable or executed by the processor. The non-transitory instructions are stored in the memory of the measurement device and can be referred to as a computer-readable storage medium and / or a non-transitory computer-readable medium. The non-transitory computer-readable medium includes all computer-readable media with the sole exception being a transient propagation signal itself. The measurement device can be a wearable device worn by the user, for example, on the wrist or finger, to measure the oxygen saturation in the user's blood.

[0017] Method 100 includes a step 110 of measuring a set of photoplethysmography (PPG) signals from light of one or more different wavelengths. In many embodiments, the set of PPG signals includes a plurality of PPG signals. Step 110 includes measuring a plurality of PPG signals from light of a plurality of different wavelengths. The PPG signal is obtained from the light absorption amount by inverting the light intensity with a photodetector after the light has passed through or been reflected from human tissue. For example, the measurement device includes an illumination element (e.g., an LED) that emits light of one or more or a plurality of wavelengths, and a photodetector that detects the light of the one or more or the plurality of wavelengths, such as light reflected from the user's skin. The one or more or the plurality of different wavelengths of light can define at least one wavelength within the range of 495 nm to 1000 nm.

[0018] Each PPG signal is measured from light of respective wavelengths or colors, such as but not limited to, for example, green, orange, red, and infrared (IR). For example, the measurement device may include four illumination elements for emitting green light, orange light, red light, and infrared light. The wavelength of the green light can be in the range of 495 - 570 nm, preferably having a peak wavelength of 536 nm. The wavelength of the orange light can be in the range of 570 - 620 nm, preferably having a peak wavelength of 610 nm. The wavelength of the red light can be in the range of 620 - 740 nm, preferably having a peak wavelength of 660 nm. The wavelength of the infrared light can be in the range of 780 - 1000 nm, preferably having a peak wavelength of 950 nm. The plurality of wavelengths can be selected from the range of 495 nm - 1000 nm, and those wavelengths can be at least 50 nm apart from each other.

[0019] Each PPG signal includes a pulsatile component and a non - pulsatile component. The pulsatile component, also known as the alternating current (AC) component, is related to the change in arterial blood volume and is synchronized with the cardiac cycle. The non - pulsatile component, also known as the direct current (DC) component, refers to the remaining portion of the PPG signal excluding the pulsatile component. The pulsatile component is superimposed on the non - pulsatile component within the PPG signal. The non - pulsatile component is related to the level of light absorption by tissue, bone, venous blood, and skin pigmentation.

[0020] It is known that as the light intensity increases, the non-pulsatile component increases in an approximately linear relationship. Further, for the same wavelength of light and the same increase in light intensity, the non-pulsatile component increases even more in people with light or non-dark skin compared to people with dark skin. This is because light absorption by the skin is affected by skin pigments such as melanin. Darker skin pigments, i.e., more melanin, absorb more light than lighter skin pigments, and less light is reflected from the skin and detected by the photodetector, thus resulting in a smaller non-pulsatile component in the PPG signal. For the same skin pigment, light absorption is different for different wavelengths of light. For example, as shown in FIG. 2A (extracted from Gonzalez-Rodriguez et al, Current Indications and New Applications of Intense Pulsed Light, Actas Dermo-Sifiliograficas, 2015), the melanin skin pigment absorbs the most green light and the least infrared light.

[0021] FIGS. 2B-2E show the relationship between the light intensity and the non-pulsatile component of light at four wavelengths: green, orange, red, and infrared. FIGS. 2B-2E each show this relationship for two skin tones: a light or non-dark skin tone and a dark skin tone. The skin tone can be classified according to the Fitzpatrick skin classification scale, with types I-III falling into the classification of non-dark skin tones and types IV-VI falling into the classification of dark skin tones. The skin tone may also be classified according to the Monk skin tone scale, with Monk01-05 falling into the classification of non-dark skin tones and Monk06-10 falling into the classification of dark skin tones. The skin tone may be classified into three or more tones, such as up to all six tones of the Fitzpatrick skin classification scale. The skin tone may also be classified according to other scales, such as the Von Luschan color scale for classifying the color of the skin.

[0022] As shown in FIGS. 2B - 2E, the non - pulsatile component is defined by the DC level measured in volts, and the light intensity is defined by the current to the lighting element measured in amperes. This relationship is established using a free - drive process in which the current is increased step - by - step and the DC level is measured at each increase step of the current. Then, the rate of increase of the DC level with respect to the light intensity is calculated as the gradient or slope. In particular, for the same wavelength, the gradient is greater in the group of non - dark - colored skin compared to the group of dark - colored skin. The gradient varies for different wavelengths due to differences in absorption ability as shown in FIG. 2A. Thus, the gradient is affected by at least two factors: skin type and wavelength.

[0023] Using pre - collected data from multiple subjects regarding gradient, wavelength, and skin tone, a first data model is constructed using one or more statistical algorithms and / or machine - learning algorithms. Preferably, the first data model is constructed using a classification algorithm and / or a regression analysis such as logistic regression. Alternatively, the first data model may be constructed using other algorithms or mathematical models such as decision trees and random forests.

[0024] Method 100 includes step 120 of calculating, for each PPG signal measured in step 110, the gradient of the non - pulsatile component of the PPG signal with respect to the light intensity of each wavelength. The gradient for each wavelength can be normalized by the gradient of one wavelength, the sum of at least two wavelengths, or the Euclidean norm of the gradients of at least two wavelengths. Method 100 includes step 130 of determining the user's skin tone from the gradients of the set of PPG signals and the first data model. Step 130 includes determining the user's skin tone from one or more gradients of the set of (one or more) PPG signals and the first data model. In many embodiments, step 130 includes determining the user's skin tone from two or more gradients of a plurality of PPG signals and the first data model.

[0025] In particular, one or more gradients from one or more different wavelengths are used to distinguish skin tones and determine the user's skin tone (i.e., dark or non - dark) from a first data model. In some embodiments, one gradient from a single - wavelength light such as, for example, green light, orange light, red light, or infrared light is used to determine the user's skin tone. Preferably, the wavelength of the light for measuring the PPG signal defines at least one of green light, orange light, red light, and infrared light. In some embodiments, multiple gradients from light of multiple wavelengths such as, for example, green light, orange light, red light, or infrared light are used to determine the user's skin tone. Preferably, the wavelength of the light for measuring the PPG signal defines at least two of green light, orange light, red light, and infrared light. In one embodiment, two gradients from orange light and infrared light are used to determine the user's skin tone. In another embodiment, three gradients from green light, orange light, and infrared light are used to determine the user's skin tone.

[0026] Tests were conducted to evaluate different wavelengths, gradients, and skin tones. Forty - nine subjects with different skin tones participated in these experiments. These forty - nine subjects included thirty - two subjects with non - dark skin, sixteen subjects with dark skin, and one subject with dark skin on the right hand and non - dark skin on the left hand. Ninety - eight skin - tone data points were obtained from both hands of the forty - nine subjects. Fifteen different combinations of one to four wavelengths of light, where each combination had one to four different wavelengths, were tested 500 times. In each iteration, fifteen subjects were randomly selected as a test data set (having 30 skin - tone data points), and the other thirty - four subjects formed a training data set (having 68 skin - tone data points). The fifteen combinations are listed below: 1. Green 2. Red 3. Infrared 4. Orange 5. Green and Red 6. Green and Infrared 7. Green and Orange 8. Red and infrared 9. Red and orange 10. Infrared and orange 11. Green, red, and infrared 12. Green, red, and orange 13. Green, infrared, and orange 14. Red, infrared, and orange 15. Green, red, infrared, and orange

[0027] Using the training dataset, a first data model was trained using logistic regression, and using the trained first data model, the performance with respect to the test dataset when determining skin tone was evaluated. In this evaluation, as shown in Figure 3A, four performance metrics were calculated: accuracy, precision, sensitivity, and specificity. Dark skin tones were defined as positive, and non - dark skin tones were defined as negative. TN means true negative and refers to the number of correct predictions of actual non - dark skin tone data points. TP means true positive and refers to the number of correct predictions of actual dark skin tone data points. FP means false positive and refers to the number of non - dark skin tone data points that were predicted as dark skin tones by the first data model. FN means false negative and refers to the number of dark skin tone data points that were predicted as non - dark skin tones by the first data model.

[0028] After 500 iterations were completed, the mean and standard deviation of each performance metric were calculated. The results for the performance metrics of accuracy, sensitivity, and precision are shown in Figures 3B - 3D. It was observed that using more different wavelengths of light achieved better performance. Single - wavelength light can produce good performance, for example, when it is orange light. However, using four different wavelengths of light would require more LEDs in the measurement device, leading to higher capacity requirements and costs. The optimal combination would be to use three different wavelengths of light to achieve good performance (i.e., a good balance and optimization of accuracy, sensitivity, and precision) while reducing the number of LEDs. It was found that the optimal combination is green, orange, and infrared light.

[0029] Furthermore, it was observed that including orange light in other colors (green, red, and infrared) improved the performance of determining skin tone. The results of green light, red light, and infrared light with and without orange light are shown in FIGS. 3E - 3H. FIG. 3H shows the F1 score, which is an overall performance metric based on the harmonic mean of the sensitivity and accuracy metrics as shown below. The F1 score improved by at least 20% when orange light was used in combination with any of green light, red light, or infrared light.

Number

[0030] The skin tone of a user depends on the melanin content of the user's skin. Different melanin contents absorb light to different extents, and more melanin absorbs more light. To distinguish skin tones, light of a suitable wavelength must penetrate to the depth of the skin where melanin is present. The suitable wavelength has a good absorption rate by melanin and varies according to the melanin content. As shown in FIG. 2A, wavelengths in the spectrum from green to orange have good penetration and are well absorbed by melanin. Therefore, these colors are suitable for distinguishing skin tones. Also, orange light penetrates deeper into the skin than green light and can reach even better to the depth where melanin is present. As reflected in the results, the use of orange light as a single - wavelength light or including orange light in combination with two or more different wavelengths of light provided good performance in determining skin tone.

[0031] The optimal combination of green, orange, and infrared wavelengths produced the best performance in determining skin tone based on gradients without using overly many LEDs and in achieving a good balance and optimization of accuracy, sensitivity, and precision. Box plots of the gradients and skin tones for each wavelength of this combination are shown in FIGS. 3I - 3K. In particular, the gradient is larger in non - dark skin tones compared to dark skin tones.

[0032] The first data model is trained using logistic regression for two skin tones (dark and non - dark), and the output of the logistic regression is the logistic score. The logistic score is the probability of being a dark skin tone. A skin tone threshold is defined to separate dark and non - dark skin tones. For example, the skin tone threshold can be defined by maximizing the F1 score.

[0033] When the user's logistic score P(dark) calculated from the first data model is greater than or equal to the skin tone threshold X, the user is classified as having a dark skin tone as shown below.

Equation

[0034] The logistic score P(dark) can be calculated using the following formula. M G is the gradient value for green light, and M O is the gradient value for orange light, and M IR is the gradient value for infrared light. a, b, c, and d are constants.

Equation

[0035] Figure 3L shows box plots of the logistics score and skin tone for combinations of green, orange, and infrared wavelengths. The dashed line represents a skin tone threshold X that can range from 0.1 - 0.7. In an exemplary experiment, this was calculated to be approximately 0.27 for combinations of green, orange, and infrared wavelengths. This value can change as more data becomes available. In particular, the logistics scores for non-dark and dark skin tones are almost completely separated, indicating that the combination of wavelengths can reliably determine whether the user's skin tone is a dark or non-dark skin tone. It is understood that the skin tone threshold will be different for each wavelength or various combinations of at least two different wavelengths. It is also understood that the first data model can also be trained to determine skin tone using light of a single wavelength.

[0036] Method 100 includes step 140 of calculating a modulation ratio from the pulsatile and non-pulsatile components of a pair of PPG signals measured from a pair of light of different wavelengths. The modulation ratio is also known as the R ratio. The pair of different wavelengths preferably defines red light and infrared light. The modulation ratio is defined as the ratio of a first quotient and a second quotient. The first quotient is derived from the pulsatile and non-pulsatile components of light of a first wavelength, such as red light for example (i.e., AC / DC). The second quotient is derived from the pulsatile and non-pulsatile components of light of a second wavelength, such as infrared light for example (i.e., AC / DC). The modulation ratio or R ratio can be defined as follows.

Number

[0037] As shown in Figure 4A, the oxygen saturation in the blood (or SpO2 level) is negatively correlated with the modulation ratio. In particular, the SpO2 level increases when the modulation ratio decreases, but their relationship is different for different skin tones. As shown in Figure 4B, for the same increase in the modulation ratio, the SpO2 level decreases slightly more in dark skin compared to non-dark skin.

[0038] Method 100 includes step 150 of selecting one from a plurality of second data models based on the user's skin tone. Method 100 includes step 160 of determining the oxygen saturation in the user's blood from the modulation ratio and the second data model selected for the user's skin tone.

[0039] To determine blood oxygen saturation, in addition to the modulation ratio, other parameters such as, for example, the first quotient or the second quotient shown below may be used. Instead, or in addition, various mathematical functions such as, for example, logarithmic functions, square roots, etc. may be applied to the modulation ratio and / or any of the parameters used in determining blood oxygen saturation.

Number

[0040] The second data model is constructed according to different skin tones such that there is a second data model for each skin tone. The second data model includes a model for dark skin tones and a model for non-dark skin tones. Using pre-collected data from a plurality of subjects on SpO2 levels, modulation ratios, and skin tones, one or more statistical algorithms and / or machine learning algorithms are used to construct the second data model. Preferably, the second data model is constructed using a classification algorithm and / or a regression analysis such as, for example, linear regression or polynomial regression. Alternatively, the second data model may be constructed using other algorithms or mathematical models such as, for example, support vector machines.

[0041] Using a training dataset of PPG signals measured from red and infrared wavelengths, a second data model was trained using regression analysis. As will be described further below, to improve the quality of the PPG signal, the PPG signal was subjected to a pulse verification process 200. The PPG signal was measured in a set of 4-second periods, and features such as modulation ratio and signal intensity were calculated for each period. The PPG signal during a certain period can be rejected if, for example, the modulation ratio is below a threshold (e.g., from 2.0 to 3.0, or preferably about 2.3), the signal intensity from the red PPG signal is below a threshold (e.g., from 0.2 to 1.5, or preferably about 0.75), and / or the signal intensity from the infrared PPG signal is below a threshold (e.g., from 0.2 to 1.5, or preferably about 0.67), when they do not meet a predetermined criterion. Then, using the qualified PPG signals, a second data model is trained with the modulation ratio as the independent variable and the SpO2 level as the dependent variable.

[0042] The training dataset was divided into dark data points and non-dark data points based on the skin tone predicted by the first data model. One second data model for dark skin tones was trained using the dark data points, and another second data model for non-dark skin tones was trained using the non-dark data points. Each second data model is selected based on the user's skin tone and can be used to measure or predict the user's SpO2 level from the modulation ratio as shown in the formula below. w, x, y, and z are coefficients derived from the second data model.

Equation

[0043] It should be understood that the first data model may be constructed using finer variations or distinctions across skin tones, such as, for example, three or more skin tones, and the second data model can be constructed according to the number of those skin tones. The finer skin tone classification can more accurately determine the user's skin tone and improve the measurement of the user's SpO2 level based on the user's skin tone. For example, the skin tone can be classified into all of Fitzpatrick scale types I-VI, or all of the 36 categories of the von Lusius color scale.

[0044] In addition to, or instead of, skin tone, the first data model may be constructed to determine the light absorbance, reflectance, and / or transmittance of other skin types or conditions and classify them into different skin tone categories. Some examples of skin types include hairy / bald skin, oily / dry skin, pigmentation on the skin (e.g., moles), or any combination thereof.

[0045] Method 100 determines the user's skin tone and then requires only a few parameters derived from the PPG signal to measure the user's blood oxygen level using the first and second data models. The second data model is selected based on the user's specific skin tone, and thus the SpO2 level predicted by the selected second data model is more accurate for that user. This addresses the problem of overestimating the SpO2 level for users with dark skin and reduces the risk of hypoxemia.

[0046] In some embodiments, method 100 includes performing a pulse verification process 200 on a pair of PPG signals for calculating a modulation ratio. The quality of the PPG signals depends on various factors such as, for example, the user's movement, ambient light, and temperature, and such factors can cause inaccurate measurements of the user's SpO2 level. This can lead to an incorrect interpretation of the SpO2 measurement and cause anxiety to the user. The pulse verification process 200 rejects pulses within the PPG signals based on a third data model when those pulses may give rise to unreliable or incorrect measurements. For example, the pulse verification process 200 rejects pulses within the PPG signals that do not meet the conditions defined in the third data model. The pulse verification process 200 ensures that the PPG signals have good quality pulses for more accurately measuring the user's SpO2 level. The third data model can be constructed from pre-collected data using one or more statistical algorithms and / or machine learning algorithms. Preferably, the third data model is constructed using a comparison of threshold values based on pulse characteristics, as well as classification algorithms and / or regression analysis such as logistic regression. Pulse characteristics include a matching difference threshold between PPG signals over a period of time, movement intensity, and signal intensity. Some of the conditions or pulse characteristics used in the third data model for the pulse verification process 200 are shown in FIG. 5A.

[0047] In one example, the conditions in the third data model include a matching difference threshold between PPG signals within a certain period (e.g., at least 4 seconds), and the PPG signal is rejected if it does not meet the matching difference threshold. For example, when a pair of PPG signals are measured from two different wavelengths such as red and infrared, the pulsation components of the pair of PPG signals should be synchronized to improve the accuracy of the modulation ratio. For example, the valleys of the pulses within each PPG signal are defined, and the corresponding pairs of valleys within that period are compared to each other. The matching difference threshold can be defined as the acceptable time limit for corresponding pairs of valleys that do not match each other. If the corresponding pairs of valleys in the PPG signal do not occur within the acceptable time limit, the pair of valleys is considered asynchronous. When the pair of valleys exceeds the matching difference threshold, this particular asynchronous pair of valleys will be rejected. Alternatively, the pulses within each PPG signal may be compared using the peaks of the pulses instead of, or in addition to, the valleys.

[0048] In one example, the conditions in the third data model include a motion intensity threshold, and pulses having a motion intensity exceeding the motion intensity threshold are rejected. The motion intensity of the pulse is related to the noise in the PPG signal that can introduce errors in SpO2 measurement. The motion intensity can be determined from the accelerometer signal. The accelerometer signal can be measured by an accelerometer module (which can measure accelerations on one, two, or three axes) within the measurement device that measures the PPG signal. When the user is moving vigorously while the PPG signal is being measured, the pulses tend to have a high motion intensity. The motion intensity threshold removes noisy pulses from the part of the PPG signal with large movement.

[0049] In one example, the conditions in the third data model include a signal intensity threshold, and pulses having a signal intensity below the signal intensity threshold are rejected. The signal intensity of the pulse is defined as the ratio of the pulsation component to the non-pulsation component of the PPG signal. A high ratio indicates a strong pulse and that there is a pulsation component sufficient to accurately calculate the modulation ratio.

[0050] Therefore, the exercise intensity threshold and the signal intensity threshold reject pulses with high noise and low signal intensity, resulting in a better PPG signal with a high signal-to-noise ratio (SNR). This improves the modulation ratio and the accuracy of subsequent SpO2 level measurements.

[0051] In one example, the conditions in the third data model include a bad pulse score threshold that represents the probability threshold that a pulse is a bad pulse or of poor quality, and pulses with a score above the bad pulse score threshold based on their morphological characteristics are rejected. More specifically, the third data model uses the bad pulse score threshold to verify the signal quality of the PPG signal based on morphological characteristics. Pulses with morphological characteristics that give a high bad pulse score are rejected.

[0052] As shown in FIG. 5A, the morphological characteristics may include a rise time that is the ratio of the interval from valley to peak to the interval from valley to valley of the pulse. The morphological characteristics may include a valley-to-valley jump that is the amplitude difference between two valleys of the pulse. The morphological characteristics may include a heart rate estimated from the interval from valley to valley of the pulse. For example, the conditions in the third data model can include a heart rate boundary, and pulses are rejected if the corresponding heart rate is outside the heart rate boundary.

[0053] The morphological characteristics may include a pulse width feature derived from the PPG signal. For example, the conditions in the third data model can include an upper pulse width threshold, and pulses having an upper pulse width (corresponding to above 50% of the systolic amplitude of the pulse) above the upper pulse width threshold are rejected. For example, the conditions in the third data model can include a lower pulse width threshold, and pulses having a lower pulse width (corresponding to below 50% of the systolic amplitude of the pulse) below the lower pulse width threshold are rejected.

[0054] Pulse width feature (PW X) represents the time interval between x% of the total height (h) of the pulse, which is the systolic amplitude of the pulse, as shown in FIG. 5A, divided by the total time interval of the pulse (from the first valley to the second valley). In the pulse verification process 200, the PW X (PW under50% ) and the PW X (PW over50% ) above 50% are used as features for classifying good pulses and bad pulses, respectively. PW 50% refers to the pulse width between points corresponding to 50% of the PPG systolic peak amplitude. PW under50% refers to the lower pulse width between points corresponding to x% of the PPG systolic peak amplitude, where this x% ranges from 0% to 50%, preferably from 20% to 30%. PW over50% refers to the upper pulse width between points corresponding to x% of the PPG systolic peak amplitude, where this x% ranges from 50% to 100%, preferably from 70% to 80%.

[0055] PW over50% can be used to identify a PPG signal damaged by unknown noise that may be caused by improper wearing of the measurement device. At x% in the range from 50% to 100%, the width of the PPG systolic peak amplitude tends to be wider in a damaged PPG signal compared to a good PPG signal. A pulse having an upper pulse width, i.e., PW over50% , that exceeds the upper pulse width threshold is classified as a bad pulse and rejected. PW under50% can be used to identify a distorted PPG signal resulting from vascular congestion caused by overly tight wearing of the measurement device. At x% in the range from 0% to 50%, the width of the PPG systolic peak amplitude tends to be narrower in a distorted PPG signal compared to a good PPG signal. A pulse having a lower pulse width, i.e., PW under50% , that is below the lower pulse width threshold is classified as a bad pulse and rejected.

[0056] An exemplary pulse verification process 200 is shown in FIG. 5B. At step 202, the peaks and valleys of the pulses within the PPG signal are defined. At step 204, the pulses within the PPG signal within the same period are compared to each other, for example, by pairs of their corresponding valleys. If the pairs of corresponding valleys in the PPG signal do not occur within an acceptable time limit, those pairs of valleys are considered asynchronous. When a pair of valleys exceeds the matching difference threshold, this particular asynchronous pair of valleys will be rejected. The matching difference threshold can be defined as 0% to 25%, preferably 10% to 25%, of the total number of pairs of valleys in the PPG signal (i.e., the sampling size).

[0057] Furthermore, step 204 may include checking whether the heart rate is within the heart rate boundary. As described above, the heart rate can be estimated from the interval from valley to valley of the pulses. If the heart rate is outside the heart rate boundary, the pulse from which the heart rate was estimated is rejected. The heart rate boundary can be, for example, a fixed standard deviation from the average heart rate determined from a sample data set of heart rate data, such as within ±10 to ±30 bpm, or preferably within ±20 bpm. Alternatively, the heart rate boundary may have a variable standard deviation from the average heart rate.

[0058] At step 206, the motion intensity of the pulses within the PPG signal is calculated from the accelerometer signal. The motion intensity represents the level of movement from the maximum magnitude of the accelerometer signal. FIG. 6A shows the effect of movement (represented by the accelerometer or ACC signal) on the PPG signal. When there is no movement, the shape of the pulses within the PPG signal can be clearly defined. When there is movement, the PPG signal is distorted with respect to the magnitude of the movement, which makes it difficult to define the peaks and valleys of the pulses.

[0059] A third data model was trained using a sample data set of PPG signals and accelerometer signals from 16 subjects. Good and bad pulses were defined based on a motion intensity threshold. The motion intensity threshold can be defined by maximizing the F1 score, which is the harmonic mean of the sensitivity and accuracy performance metrics of the third data model with respect to the strength of movement. The motion intensity threshold can be in the range of 0.01 - 0.1, such as 0.018 for example, in this sample data set. As shown in the box plot of FIG. 6B, good pulses are associated with low motion intensity.

[0060] In step 208, the signal strength of the pulses in the PPG signal is calculated. Preferably, the signal strength of the infrared PPG signal is used. The signal strength determines whether the magnitude of the pulse is large enough for the PPG signal to be considered clearly defined.

[0061] FIG. 7A shows exemplary PPG signals measured from different sources under low to no motion. The first and second PPG signals are measured from a human user, and the third PPG signal is measured from an inanimate object. As shown in the box plot of FIG. 7B, the signal strength of the third PPG signal (around 0.009) is relatively low compared to the first and second PPG signals (around 0.55 and around 0.01 - 0.015 respectively). Although both PPG signals are measured from a human user, the signal strength of the second PPG signal is lower than that of the first PPG signal. This is because the signal strength is affected by factors such as loose attachment of the measurement device to the user's body, body temperature, and skin tone. As shown in the second PPG signal of FIG. 7A, even if the second PPG signal is measured under low to no motion, the low signal strength can cause noise and distort the pulse shape. The signal strength threshold can be defined based on the minimum signal strength that still gives an acceptable pulse shape where peaks and valleys can still be clearly defined. The signal strength threshold can be in the range of 0.01 - 0.05, such as 0.013 for example, in this sample data set.

[0062] Furthermore, in step 208, morphological features of the PPG signal are calculated. For example, the morphological features include the rise time and valley-to-valley jump of each pulse in the red and infrared PPG signals, and the valley-to-valley jump of each pulse in the red PPG signal.

[0063] Figures 8A - 8C show the distribution of these morphological features into good pulses and bad pulses. In particular, for bad pulses, the standard deviation is wider, indicating that bad pulses are further dispersed by noise randomness. A third data model is trained by logistic regression using a sample data set of these morphological features. The bad pulse score threshold can be defined by maximizing the F1 score based on similar performance metrics for the bad pulse score. The bad pulse score threshold can be in the range of 0.1 - 0.5, for example 0.14, in this sample data set.

[0064] Figures 8D and 8E show the mean absolute error (MAE) of SpO2 in different ranges of the pulse width features PW over50 and PW under50 . The MAE is larger in two scenarios: when PW over50 is higher than a specific value and when PW under50 is lower than a specific value. These trends are observed in both the red PPG signal and the infrared PPG signal, and this data is used to train a third data model. Pulses with PW over50 higher than the threshold or PW under50 lower than the threshold are classified as bad pulses and rejected. In this example, the thresholds for PW over50 and PW under50 are 0.54 and 0.42 respectively.

[0065] In step 210, the bad pulse score of each pulse in the PPG signal is calculated from the morphological features and the third data model. In step 212, each pulse is looped through and enters a pulse classification process 220 for classifying the pulse as good or bad.

[0066] In the pulse classification process 220, for each pulse, step 222 compares the exercise intensity with an exercise intensity threshold. If the exercise intensity is below the threshold, step 222 proceeds to step 224. If the exercise intensity is above the threshold, step 222 is a failure, and that pulse is classified as a bad pulse at step 230. Step 224 compares the signal intensity with a signal intensity threshold. If the signal intensity is above the threshold, step 224 proceeds to step 226. If the signal intensity is below the threshold, step 224 is a failure, and that pulse is classified as a bad pulse at step 230. Step 226 compares the bad pulse score with a bad pulse score threshold. If the bad pulse score is below the score threshold, step 226 proceeds to step 228a. If the bad pulse score is above the score threshold, step 226 is a failure, and that pulse is classified as a bad pulse at step 230. Step 228a compares the pulse width (PW over50% ) above 50% of the amplitude with a threshold. If PW over50% is below the threshold, step 228a proceeds to step 228b. If PW over50% is above the threshold, step 228a is a failure, and that pulse is classified as a bad pulse at step 230. Step 228b compares the pulse width (PW under50% ) below 50% of the amplitude with a threshold. If PW under50% is below the threshold, step 228b is a failure, and that pulse is classified as a bad pulse at step 230. If PW under50% is above the threshold, step 228b proceeds to step 232, and that pulse is classified as a good pulse. It should be understood that steps 222, 224, 226, 228a, and 228b may be performed in any order.

[0067] The pulse verification process 200, therefore, retains the good pulses within the PPG signal, and the refined PPG signal is used to calculate a modulation ratio and then to measure the SpO2 level. In some embodiments, the measured SpO2 level is used to further verify the PPG signal in the pulse verification process 200. More specifically, a measurement device incorporating the method 100 is used to measure the heart rate and SpO2 level from the PPG signal. The measured heart rate and the measured SpO2 level are compared with a reference heart rate and a reference SpO2 level measured from a reference device, such as a finger pulse oximeter. The measurement device and the reference device are communicable with each other to perform this comparison.

[0068] For each period measured over a period, such as a 4-second period for example, a heart rate error is calculated by the absolute difference between the measured heart rate and the reference heart rate, and an SpO2 error is calculated by the absolute difference between the measured SpO2 level and the reference SpO2 level. If the SpO2 error is at most 2 standard deviations and the heart rate error is at most 30 bpm, the pulses of the PPG signal within the period are classified as good pulses. If the SpO2 error is greater than 2 standard deviations or the heart rate error is greater than 30 bpm, the pulses of the PPG signal within the period are classified as bad pulses.

[0069] Using the SpO2 error, the heart rate boundary may be optimized and made adaptable to the SpO2 measurement value. FIG. 9 shows an exemplary iterative process 250 for optimizing the heart rate boundary. At step 252, a sample data set is input into the iterative process 250. The sample data set includes heart rate data from 13,112 periods of 4 seconds each. At step 254, for each period, the heart rate boundary is defined as within the standard deviation from the average heart rate (e.g., within 10 - 30 bpm, etc.). At step 256, if the corresponding heart rate is outside the heart rate boundary, the pulse is rejected. At step 258, after rejecting the bad pulses, the modulation ratio is calculated using the resulting PPG signal, and the SpO2 level is measured. The SpO2 error is also measured compared to the reference SpO2 level. At step 260, the standard deviation is changed, and steps 254, 256, and 258 are repeatedly iterated based on the changed standard deviation. The standard deviation is changed until an optimized standard deviation can be found at step 262. The optimized standard deviation is selected based on the lowest SpO2 error and the smallest number of rejected pulses.

[0070] The pulse verification process 200 thus helps to exclude bad or low-quality pulses from the PPG signal, calculate the modulation ratio, and then improve the overall quality of the PPG signal used to measure or predict the SpO2 level from the selected second data model. An exemplary embodiment of method 100 including the pulse verification process 200 is shown in FIG. 10 as method 300 for measuring the SpO2 level in a user's blood.

[0071] In step 302, gradients from green, orange, and infrared PPG signals are calculated. In step 304, a logistics score is calculated from those gradients and a first data model. In step 306, the logistics score is compared with a skin tone threshold. If the logistics score is below the skin tone threshold, the user's skin tone is predicted to be non-dark (step 308). If the logistics score is equal to or higher than the skin tone threshold, the user's skin tone is predicted to be dark (step 310).

[0072] In step 312, for verification by the pulse verification process 200, pulses within the PPG signal are detected within a period of at least 4 seconds. Step 314 checks whether the heart rate is within the heart rate boundary. If the heart rate is outside the boundary, step 314 fails, and the pulse is rejected in step 316. If the heart rate is within the boundary, step 314 proceeds to step 318 which checks for matching between the red PPG signal and the infrared PPG signal. If the corresponding pair of valleys in the PPG signal do not occur within an acceptable time limit, those pairs of valleys are considered to be above the matching difference threshold and asynchronous, step 318 fails, and the PPG signal during that period is rejected in step 316. If the corresponding pair of valleys in the PPG signal occur within an acceptable time limit, that is, below the matching difference threshold, those pairs of valleys are considered to be synchronous, and step 318 proceeds to step 320 which compares the exercise intensity of each pulse with the exercise intensity threshold. If the exercise intensity is above the exercise intensity threshold, step 320 fails, and the pulse is rejected in step 316. If the exercise intensity is below the exercise intensity threshold, step 320 proceeds to step 322 which compares the signal intensity of each pulse with the signal intensity threshold. If the signal intensity is below the signal intensity threshold, step 322 fails, and the pulse is rejected in step 316. If the signal intensity exceeds the signal intensity threshold, step 322 proceeds to step 324 which compares the bad pulse score of each pulse with the bad pulse score threshold. If the bad pulse score is above the bad pulse score threshold, step 324 fails, and the pulse is rejected in step 316. If the probability of the bad pulse score being below the bad pulse score threshold, step 324 proceeds to step 325a. PW over50% If it exceeds the threshold, step 325a fails, and the pulse is rejected in step 316. PW over50% If it is below the threshold, step 325a proceeds to step 325b. PW under50% If it is below the threshold, step 325b fails, and the pulse is rejected in step 316. PW under50%If it exceeds the threshold, the pulse is classified as a good pulse, and step 325b proceeds to step 326. It should be understood that steps 314, 318, 320, 322, 324, 325a, and 325b may be executed in any order.

[0073] Step 326 determines whether there are pulses remaining in the PPG signal after the pulse verification process 200. If no pulses remain, step 326 proceeds to step 328, and the SpO2 level for that period cannot be measured. If pulses remain, they are pulses of good quality, and step 326 proceeds to step 330. Step 330 checks whether the modulation ratio of the PPG signal is above a threshold. If not, the PPG signal is rejected, and the SpO2 level cannot be measured (step 328). If so, step 330 proceeds to step 332, which checks whether the signal intensities from the red and infrared PPG signals are below a threshold. If not, the PPG signal is rejected, and the SpO2 level cannot be measured (step 328). If so, step 332 proceeds to step 334.

[0074] Step 334 selects a second data model based on the user's skin tone determined in step 306. If the user's skin tone is dark, a second data model for dark skin tones is selected and used to predict the SpO2 level for that period from the modulation ratio (step 336). If the user's skin tone is non-dark, a second data model for non-dark skin tones is selected and used to predict the SpO2 level for that period from the modulation ratio (step 338).

[0075] Method 300 was tested using 98 skin tone data points from the 49 subjects described above. A combination of green light, orange light, and infrared light was tested for 1000 iterations. In each iteration, 15 subjects (i.e., 30 skin tone data points) were randomly selected as the test data set, and the other 34 subjects (i.e., 68 skin tone data points) were used to form the training data set. Using the training data set, a first data model was trained using logistic regression, and the performance of the trained first data model was evaluated against the test data set when determining skin tone. After completing 1000 iterations, the average values and standard deviations of four performance metrics from the test data set were calculated as follows and their distributions are shown in FIGS. 11A and 11B.

[0076] Accuracy: average 0.86, standard deviation 0.07 Precision: average 0.82, standard deviation 0.16 Sensitivity: average 0.76, standard deviation 0.17 Specificity: average 0.91, standard deviation 0.09

[0077] The pulse verification process 200 was able to classify good pulses and bad pulses within the PPG signal. The performance of the pulse verification process 200 was evaluated against a sample of 17,549 pulses, as shown in the matrix of FIG. 11C. Good pulses are defined as "0" and bad pulses are defined as "1". The good and bad pulses predicted by the pulse verification process 200 were compared to the target criteria of independently classified good and bad pulses. In one example, the target criteria for good and bad pulses are obtained by visual observation. Good pulses have a distinct systolic peak and / or diastolic peak, regardless of whether the systolic wave and diastolic wave are clearly defined. Bad pulses have indistinguishable systolic and diastolic waves and / or an inconsistent pulse baseline. In another example, the target criteria for good and bad pulses are obtained with the aid of a finger pulse oximeter. The predicted values of the SpO2 error and HR error are compared to the reference values (from the finger pulse oximeter), and if the difference in the comparison is small (e.g., SpO2 error ±2SD, HR error ±30 bpm), the pulse is defined as good. Approximately 86.4% and 6.6% of the pulses were correctly predicted as good pulses and bad pulses, respectively. The performance metrics are 93.0% accuracy, 55.7% precision, 79.4% sensitivity, and 94.3% specificity.

[0078] Method 300 was used on 33 subjects to predict their SpO2 levels based on their skin tone. The predicted SpO2 levels were compared to reference SpO2 levels measured by a reference device such as a finger pulse oximeter. The performance of the SpO2 prediction by Method 300 was tested in 1000 iterations. In each iteration, 10 skin tone data points were randomly selected as the test data set, and the other 23 skin tone data points were used to form a training data set for training a second data model using regression analysis. The training data set was split into two subsets by skin tone. One was for training a second data model for dark skin tones, and the other was for training a second data model for non-dark skin tones. The test data set was similarly split into two subsets by skin tone. After completing 1000 iterations, the mean value and standard deviation of the root mean square error (RMSE) between the predicted SpO2 levels and the reference SpO2 levels from the test data set were calculated to be 2.73 and 0.32, respectively. The distribution of the RMSE is shown in FIG. 11D. The RMSE is within the 3.5% tolerance defined by the US Food and Drug Administration.

[0079] In the foregoing detailed description, embodiments of the present disclosure regarding methods and devices for determining skin tone and measuring oxygen saturation in blood have been described with reference to the provided drawings. The description of the various embodiments herein is not intended to evoke or be limited to only specific or particular representations of the present disclosure, but is merely intended to show non-limiting examples of the present disclosure. The present disclosure serves to address at least one of the problems and issues referred to in relation to the prior art. Although only some embodiments of the present disclosure are disclosed herein, it will be apparent to those skilled in the art in view of this disclosure that various modifications and / or changes can be made to the disclosed embodiments without departing from the scope of the present disclosure. Accordingly, the scope of the present disclosure and the scope of the following claims are not limited to the embodiments described herein.

Claims

1. A computerized method for measuring the oxygen saturation level in a user's blood, A set of photoplethysmography (PPG) signals is measured from one or more different wavelengths of light, and each PPG signal is measured from its respective wavelength and has pulsating and non-pulsating components. For each PPG signal, calculate the gradient of the non-pulsating component of the PPG signal with respect to the light intensity of each wavelength. The user's skin tone is determined from one or more gradients of the set of PPG signals and a first data model. The modulation ratio is calculated from the pulsating and non-pulsating components of a pair of PPG signals measured from a pair of light of different wavelengths. Based on the user's skin tone, one of several second data models is selected. The oxygen saturation level in the user's blood is determined from the modulation ratio and the second data model selected for the user's skin tone. A method having the following characteristics.

2. The set of PPG signals has multiple PPG signals, and the method is The multiple PPG signals are measured from light of multiple different wavelengths. The user's skin tone is determined from the gradients of one or more of the plurality of PPG signals and the first data model. The method according to claim 1, wherein the method is as follows:

3. The method according to claim 1 or 2, wherein the one or more different wavelengths of light define at least two wavelengths in the range of 495 nm to 1000 nm.

4. The method according to claim 3, wherein the at least two wavelengths are separated by at least 50 nm from each other.

5. The method according to claim 1, wherein the one or more different wavelengths of light define at least two of green light, orange light, red light, and infrared light.

6. The method according to claim 1, wherein the first data model and the second data model are constructed using one or more machine learning algorithms.

7. The method according to claim 6, wherein the machine learning algorithm comprises a classification algorithm and / or regression analysis.

8. The method according to claim 1, wherein the second data model comprises a model for dark skin tones and a model for non-dark skin tones.

9. The method according to claim 1, further comprising rejecting the pair of PPG signals when the modulation ratio falls below a predetermined threshold.

10. The method according to claim 1, further comprising rejecting the pair of PPG signals if the signal intensity of one or both PPG signals falls below a predetermined threshold.

11. The method according to claim 1, wherein the pair of wavelengths of light define red light and infrared light.

12. The method according to claim 1, wherein a pulse verification process is performed on the pair of PPG signals for calculating the modulation ratio, the pulse verification process being for rejecting pulses in the PPG signals that do not satisfy the conditions defined in a third data model.

13. The method according to claim 12, wherein the condition in the third data model has a matching difference threshold between the PPG signals over a certain period, and the PPG signal is rejected if the matching difference threshold is not met.

14. The method according to claim 12 or 13, wherein the condition in the third data model has an exercise intensity threshold, and pulses having an exercise intensity exceeding the exercise intensity threshold are rejected.

15. The method according to claim 14, further comprising measuring an accelerometer signal to determine the exercise intensity.

16. The method according to claim 12, wherein the condition in the third data model has a signal intensity threshold, and pulses having a signal intensity below the signal intensity threshold are rejected.

17. The method according to claim 12, wherein the conditions in the third data model have a bad pulse score threshold, and pulses with a score exceeding the bad pulse score threshold are rejected based on their morphological characteristics.

18. The method according to claim 12, wherein the condition in the third data model has a heart rate boundary, and a pulse is rejected when the corresponding heart rate is outside the heart rate boundary.

19. The method according to claim 18, further comprising optimizing the heart rate boundary in an iterative process.

20. The method according to claim 12, wherein the condition in the third data model has an upper pulse width threshold, and pulses having an upper pulse width exceeding the upper pulse width threshold are rejected, and the upper pulse width corresponds to more than 50% of the systolic amplitude of the pulse.

21. The method according to claim 12, wherein the condition in the third data model has a lower pulse width threshold, and pulses having a lower pulse width below the lower pulse width threshold are rejected, and the lower pulse width corresponds to less than 50% of the systolic amplitude of the pulse.

22. The method according to claim 12, wherein the third data model is constructed using one or more machine learning algorithms.

23. The method according to claim 22, wherein the machine learning algorithm has a comparison of pulse features and a threshold based on logistic regression.

24. A computerized method for determining a user's skin tone, The current supplied to the lighting element is increased in a series of increasing steps to control the lighting element so that it emits light of one or more different wavelengths. A set of photoplethysmography (PPG) signals is measured from one or more different wavelengths of light, and each PPG signal is measured from its respective wavelength and has a pulsating component and a non-pulsating component for each of the multiple incrementing steps. For each PPG signal, the gradient of the non-pulsating component of the PPG signal with respect to the light intensity of each wavelength is calculated over the plurality of increasing steps, and the light intensity for each increasing step is defined by the current supplied to the illumination element in that increasing step. The user's skin tone is determined from one or more gradients of the set of PPG signals and a first data model. A method having the following characteristics.

25. The set of PPG signals has multiple PPG signals, and the method is The multiple PPG signals are measured from light of multiple different wavelengths. The user's skin tone is determined from the gradients of one or more of the plurality of PPG signals and the first data model. The method according to claim 24, wherein the above is achieved.

26. The method according to claim 24 or 25, wherein the one or more different wavelengths of light define at least one wavelength in the range of 495 nm to 1000 nm.

27. The method according to claim 26, wherein the one or more different wavelengths of light define at least two wavelengths in the range of 495 nm to 1000 nm.

28. The method according to claim 27, wherein the at least two wavelengths are separated from each other by at least 50 nm.

29. The method according to claim 24, wherein the one or more different wavelengths of light define at least one of green light, orange light, red light, and infrared light.

30. A measuring device for measuring the oxygen saturation level in a user's blood, One or more illuminating elements that emit light of one or more different wavelengths, One or more photodetectors that detect light of one or more different wavelengths, A processor configured to perform the computerized method described in claim 1, A measuring device having the following features.

31. A non-temporary computer-readable storage medium storing computer-readable instructions, wherein, when executed, the computer-readable instructions cause a processor to execute the computerized method for measuring the oxygen saturation level in a user's blood as described in claim 1.

32. A measuring device that determines the user's skin tone, One or more illuminating elements that emit light of one or more different wavelengths, One or more photodetectors that detect light of one or more different wavelengths, A processor configured to perform the computerized method described in claim 24, A measuring device having the following features.

33. A non-temporary computer-readable storage medium storing computer-readable instructions, wherein, when executed, the computer-readable instructions cause a processor to execute a computerized method for determining a user's skin tone as described in claim 24.

Images