Pigments in various human skin types could affect signal quality in photoplethysmography (PPG) and could even make physiological measurements invalid. The quality of PPG-based in vivo physiological monitoring is attributed to melanin concentration of skin and its related pigments of skin types [1
]. Thereby, melanin in human skin is known to highly attenuate incident light with relevant wavelength illuminations [2
]. Some researchers reported anecdotal errors (3–5%) in black skin [3
]. In addition, modelling and simulations were conducted and errors due to various pigments were reported [3
]. Ink on skin and nail polish could cause errors during measurements [5
]. Four different wavelength illuminations have attempted to optimize the measurements of heart rate (HR) by using a customized PPG setup on 23 healthy subjects with various skin types on PPG system [6
]. The results showed that the 520 nm wavelength (green) illumination provided a significantly greater (p
< 0.001) ability to detect HR. Increasing levels of melanin, or darker skin type (Type V) showed a decreasing trend with insignificant change (p
< 0.067). Moreover, it has been widely confirmed that other forms of medium or material, i.e., fingerprinting ink and henna, could bring some issues to optical absorption and thus produce errors during the measurements. Some kinds of wrist tattoos could also interfere with PPG sensors on the back of smart watches, resulting in errant watch functionalities with users [7
The skin of the human body is comprised of three main layers: epidermis, dermis and subcutaneous. Melanin only plays an essential role in skin color and is responsible for a wide variety of skin color complexions. At present, Von Luschan’s chromatic scale (VLCS) and Fitzpatrick scale (FPS) are applied in skin classification. Usually, VLCS [8
] is applied to establish racial classifications of populations based upon skin colors. In contrast, FPS is considered [9
] for classification of individual skin type, as introduced in 1975, to describe sun-tanning behavior. FPS distinguishes several separate skin tones that culturally fall under “white,” or Caucasian, and does not make an adequate distinction on the darker side of human skin color gradient [10
]. Hence, VLCS was applied in this study to better describe the actual varieties of skin colors within these ethnic group people. The majority of skin types, in corresponding regions, are defined in Table 1
Meanwhile, it would be useful to reveal human tissue optical properties in order to better understand the interaction of illumination to tissue with physiological variations. Skin absorption rates are usually various when the skin is illuminated by different wavelength illuminations. Figure 1
illustrates an optical window of human tissue in which the illumination is greatly absorbed by melanin within the ultraviolet band (10–400 nm), whereas the water substance is strongly absorbed in the range of longer wavelength. Generally, the optical window (405–1064 nm) is often chosen for opto-physiological measurements [11
], such as a typical application of pulse oximeter.
In addition, the illumination wavelength of PPG is usually determined based on the absorption of oxyhemoglobin (Hb) and deoxyhemoglobin (HbO2
) associated with blood volume changes in peripheral and capillary arterial vessels. Namely, the illumination associated with specific wavelengths to skin with rich peripheral blood vessels are preferable to surface skin layers where there is no arterial blood or scattering. Table 2
shows the wavelengths of optical radiation reaching a certain penetration depth.
Various skin pigments are highly correlated with spectrometric illumination on skin surface, thus impacting the quality of the PPG measurements as shown in Table 2
. In addition, different optical radiation wavelengths can reach different penetration depths of skin tissue. Hence, this study aims to investigate measurement effectiveness of multi-wavelength illumination based on opto-electronic patch sensor (OEPS) against skin pigments using new electronic architecture with auto adaptive adjustment of signals.
The evaluation of the HR readings between two measurement techniques, i.e., OEPS and three-lead resting ECG (AT-10 Plus, Schiller UK Ltd., Bellshill, UK) were performed using unpaired t
-test. Pearson’s correlation analysis was also used to correlate quantitative variables (r
≥ 0.98), as an indicator of two techniques evolving in parallel, as shown in Figure 7
. The test showed that there was no significant difference since there is a probability value of (p
= 0.99) with a condition of (p
The Bland–Altman method was used to compare the values of the HR obtained between the OEPS and the commercial ECG devices. Figure 8
indicates the bias B: 0.04 bpm, standard deviation SD = 2.37 bpm, lower and upper limits of agreement, −4.60 bpm and +4.68 bpm respectively.
Thirty-three healthy male subjects, with five different skin types divided in four ethnic groups, were selected to participate in the implementation of the experimental protocol. The average value range (AVR, bpm) and the standard deviation range (SDR, bpm) of these five skin types of subjects, during the four activities, are summarized in Table 3
. In resting, the AVRs of HR were from 57 to 94 and SDRs were from 1.0 to 5.9. Whilst running, the AVRs of HR were increased from 98 to 170 and the SDRs from 3.1 to 21.9 bpm. The largest value of SDR was at 21.9 bpm.
The mean HRs and the lower and upper limits of HR in the four exercise conditions (resting, walking, jogging and running) were obtained referring to the skin types as shown in Figure 9
The results through the two-way ANOVA execution displays that p
1 ≈ 0 (for activities) and p
2 = 0.09 (for skin types) as obtained from the list in Table 4
presents signals from one subject before and after auto adaption gain for different exercises, i.e., resting, walking and jogging.
The outcome achieved from the implementation of the designated experimental protocol has demonstrated the OEPS, with its associated electronics and algorithms. This is also able to meet the specified skin type through comparison between commercial three-lead resting ECG and the OEPS developed in the lab. The optimal illumination to a designated skin type can be selected by a desired spectral wavelength from 525 nm to 870 nm, together with four channels with 16 LEDs, through an auto-adaptive adjustment. Undoubtedly, multi-wavelength illuminations, for instance having more LEDs attached to the OEPS, could improve the performance of opto-physiological monitoring. A better performance microcontroller, i.e., dsPIC33FJ256MC710, could be ideal to control these multiplexed LED drivers and to demultiplex the signals captured from the PD. However, this could increase the complexity of the OEPS system. Thus, a better solution to optimize the trade-off between performance and cost is developing a forthcoming OEPS architecture design. Figure 7
presents the HR measurement between the OEPS and the three-lead resting ECG during resting in a higher correlation (r:
0.982). As depicted in the Bland–Altman plot in Figure 8
, the HR outputs recorded at resting are most likely to be in the acceptable range of B ±
in the HR difference between the three-lead resting ECG and the OEPS. From Figure 7
and Figure 8
it can be derived that the HR measurements obtained from the commercial ECG device and the OEPS are not only compatible but satisfactory as well [18
As presented in Figure 9
, the HRs among the different five types of skin during the four types of exercises, were measured by means of the OEPS. The largest value of SDR during resting and walking is 7.2 bpm. While jogging and running, the HRs are fluctuated significantly and the largest SDR reaches up to 21.9 bpm. The main reasons for the increase in the SDR value may be caused by excessive strain of the OEPS sensor applied on the subject palm as well as the electrical noise from the PD during the exercises. In addition, the HRs on different people could be varied due to individual blood circulations. The two-way ANOVA showed that the activities have a notable factor because p1
≈ 0 (F
= 323.2) which is less than p
= 0.01, whilst the different types of skin did not have a significant factor because p2
= 0.09 (F
= 3.0) as it is more than p
Referring to the auto adaption effect shown in Figure 10
a–f, the intensities of the PPG signals appear to be higher with the auto gain applied compared to those without the auto gain. As depicted in Figure 10
, the auto gain is able to improve the quality of the signal-to-noise ratio (SNR) resulting in more precise and easy-to-obtain HR measurements. Meanwhile, starting from the resting exercise to the jogging one, the intensities of the PPG signals became higher over time. As the intensity of the physical activity in the jogging exercise increased, the intensity of the PPG signals corresponding to the illuminations of yellow, red and infrared LEDs was also increased significantly, whilst the intensity of the PPG signals from the green illumination remained stable. However, the motion artifact occurred predominantly with the increase of the physical activity resulting in imprecise HR measurements contrary to resting or moderate activity.
Fallow, B.A. et al. [6
] mentioned that four-different-wavelength illumination was used in the assessment of HR detection (Blue 470 nm, Green 520 nm, Red 630 nm, and Infrared 870 nm). The results showed that green illumination has greater modulation than the other wavelength illuminations regardless of skin types. As a matter of fact, the same conclusion was reached in the utilization of the OEPS as shown in Figure 10
. Thus, the HRs were mainly attained by the green wavelength illumination in the OEPS.
A multi-wavelength opto-electronic patch sensor (OEPS), through the implementation of a designated exercise protocol, has been proved to effectively detect physiological signals against skin pigments. The auto adaptation module has been realized to optimally select a suitable illumination for the right type of skin. The study shows the approach of acquiring and processing physiological signals that could be effectively applied in real time, such as sending signals wirelessly to a PC monitor or a server on a receiving end user. With a comparison measurement, the compatibility of HRs between the ECG and the OEPS was proven to be in an acceptable range of the HR difference using the three-lead ECG and the OEPS. Specifically, the outcome of the experimental protocol, with the participation of 33 healthy male subjects, demonstrates the OEPS ability to measure HR effectively regardless of the skin types during various physical activities. The OEPS, with multi-wavelength illuminations, could be extended to the scope of sport physiological monitoring application, and could be applied in various ethnic groups for cost-effective health monitoring and assessment. Furthermore, the OEPS could be consolidated to already available wearable and smart devices for in-line and real-time monitoring and assessment. Nevertheless, several challenges such as different OEPS measurement locations on the skin with relevantly rich peripheral blood vessels, optimum wavelength illuminations, motion artifact, and even electronic noises, are to be addressed in the upcoming studies.