Toward an Accurate IR Remote Sensing of Body Temperature Radiometer Based on a Novel IR Sensing System Dubbed Digital TMOS

Abstract

A novel uncooled thermal sensor based on a suspended transistor, fabricated in standard CMOS-SOI process, and released by dry etching, dubbed Digital TMOS, has been developed. Using the transistor as the sensing element has advantages in terms of internal gain, low power, low-cost technology, and high temperature sensitivity. A two channel radiometer, based on the new nano-metric CMOS-SOI-NEMS Technology, enables remote temperature sensing as well as emissivity sensing of the forehead and body temperatures of people, with high accuracy and high resolution. Body temperature is an indicator of human physiological activity and health, especially in pediatrics, surgery, and general emergency departments. This was already recognized in past pandemics such as SARS, EBOLA, and Chicken Flu. Nowadays, with the spread of COVID-19, forehead temperature measurements are used widely to screen people for the illness. Measuring the temperature of the forehead using remote sensing is safe and convenient and there are a large number of available commercial instruments, but studies show that the measurements are not accurate. The surface emissivity of an object has the most significant effect on the measured temperature by IR remote sensing. This work describes the achievements towards high–performance, low-cost, low power, mobile radiometry, to rapidly screen for fever to identify victims of the coronavirus (COVID-19). The main two aspects of the innovation of this study are the use of the new thermal sensor for measurements and the extensive modeling of this sensor.

1. Introduction

1.1. The Role of IR Remote Temperature Measurements

IR Radiation thermometry is attractive in many challenging temperature measurement situations because it is a non-contact, non-intrusive, and fast technique. Thermal radiation is governed by the fundamental physical laws established over one hundred years ago by Kirchhoff, Stefan, Boltzmann, Wien, and Planck. These laws directly link emitted blackbody radiation, totally or spectrally resolved, to the thermodynamic temperature of the emitting source. Actual practical measurements by radiation thermometry, however, are prone to several uncertainties associated with, for example, surface emissivity and environmental effects such as ambient temperature and reflected ambient radiation [1,2].

1.2. The Role of Emissivity

The recent COVID-19 pandemic has increased the need and motivation for accurate and low-cost thermometers that detect human body temperature with a high accuracy of 0.1 °C. However, a deeper understanding of the physics of the topic under study teaches us that this accuracy cannot be achieved without determining the surface emissivity as well. This defines the main research question of the present study: how we measure accurately the surface emissivity, as well as the surface temperature.

Emissivity has a coefficient from 0 to 1 that is the ratio of the emitted energy of an object to that of a theoretical blackbody at the same temperature. All infrared (IR) thermometry is based on the concept of the blackbody, which eventually leads mathematically to the concept of emissivity. Planck’s law provides the only theoretically rigorous link between the temperature of a blackbody and a radiated energy of precise quantifiable characteristics. However, emissivity is an elusive parameter. It makes sense only for gray bodies; a gray body is an imperfect blackbody, i.e., a physical object that partially absorbs incident electromagnetic radiation. The ratio of a gray body’s thermal radiation to a blackbody’s thermal radiation at the same temperature is called the emissivity of the gray body. The emissivity of the gray body does not depend on the wavelength for a given range of optical bandpass [1].

1.3. The Human Body Temperature and Emissivity

Based on numerous studies, it is now established that many factors affect the body temperature and emissivity, in the presence of extrinsic (i.e., hot environments) or intrinsic factors (i.e., pathology, exercise), such as: ambient temperature; exercise; sweat; blood pressure; superficial blood flow; heat conduction from deeper tissues (including muscle); and heat loss across the skin’s surface. The skin is also the site of reciprocal heat transfer between the human body and the external environment. The extent to which heat transfer occurs is largely dependent on environmental conditions, such as ambient temperature, water vapor, and the thermal properties of human skin [3,4,5,6,7,8]. The following facts are now clear:

The human body controls core body temperature within a tight band typically between 36 and 38 °C, despite varying ambient temperatures. Thermoreceptors located within the hypothalamus, spinal cord, skin, and some abdominal organs monitor temperature changes, and work via negative feedback to initiate autonomic mechanisms to either conserve (via shivering and vasoconstriction) or lose (via sweating and vasodilation) body heat [9].

In summary, the forehead skin emissivity is mainly determined by the water contents. In the optical bandpass of ~5–14 µm it is around 0.98, regardless of the ethnicity or gender. Living people’s temperature may change only in the range of 32–42 °C; the emissivity may change more or less by ±0.005 around 0.98 (see Appendix B).

2. The Building Block of the Radiometer under Study

2.1. The Digital TMOS

In recent years, there have been tremendous developments in instrumentation. For instance, infrared focal plane arrays can now produce images with a spatial resolution of the order of 10 mm with a temperature resolution of 0.01 K. However, these imagers cannot detect the human body temperature correctly by remote sensing. As explained above, a radiometer that detects both temperature and emissivity is required. Therefore, a two-channel sensing system is required. The preferred IR sensor for the radiometer should be linear, reproducible, sensitive to small temperature changes, accurate, and manufactured with a mature technology that requires low power during operation (see Section 3).

The TMOS, short for Thermal-MOS, is a thermal sensor based on a thermally insulated MOSFET transistor. The TMOS is manufactured by the CMOS-SOI and MEMS process. A schematic description of the analog TMOS is shown in Figure 1.

The micro-machined thermally insulated transistor has very low thermal mass and very low thermal conductivity. The absorbed photons increase the TMOS temperature and modify the current–voltage characteristics, and since the transistor is operated at a subthreshold region, its I-V characteristics are highly dependent on temperature (exponentially), therefore a highly sensitive sensor is achieved [10,11,12,13,14,15,16,17].

The Digital TMOS is described in Figure 2.

The sensor is based on mosaic design, and it is differential: it contains a blind sensor with identical design, which is covered by mirror. Therefore, the DC current is canceled by differential reading. The electrical signal is read and processed on the ASIC.

The packaged device is available commercially [18] and is shown in Figure 3.

The two-channel radiometer under study contains two digital TMOS units, each TMOS is covered with a different filter, first, a generic filter of 5.5–13.5 µm (wide bandpass filter), and the second one with an external filter with a bandwidth of 8–14 µm (narrow bandpass filter). The device is shown in Figure 4.

2.2. The Radiometer under Study

2.2.1. With No Optics

The detector (without lens) is placed in front of a forehead (Figure 5). In the case of calibration, the detector views only a large and uniform emitting surface of a blackbody, instead of a forehead, without any chopping.

Parameters:

• H—distance between a target (forehead/blackbody) and a detector;
• A_D—detector area;
• A_T—target area;
• t_filter—optical filter transmittance in a wavelengths range λ1–λ2 or λ3–λ4;
• $\theta /2$—Half Field-of-View (FOV);
• ε–object target emissivity;
• W_λ1–λ2 (T_BB)—blackbody emitting power according to Planck’s Law in a given optical bandpass;
• η—absorption coefficient

The power incident absorbed by the detector from the target is equal to the object-emitted power and the reflected power from the ambient environment by the object:

$$\begin{array}{c}{P}_{\lambda 1-\lambda 2}\left(\epsilon , T_{obj},T_{amb}\right)=\underset{emitted by the object}{\underbrace{{\epsilon }_{obj}{A}_D\cdot W_{\lambda 1-\lambda 2}\left(T_{obj}\right)\cdot t_{filter}\cdot \eta \cdot {\sin }^2\left(\frac{\theta }{2}\right)}}+\\ \underset{reflected}{\underbrace{\left(1-{\epsilon }_{obj}\right)\cdot A_D\cdot W_{\lambda 1-\lambda 2}\left(T_{amb}\right)\cdot t_{filter}\cdot \eta \cdot {\sin }^2\left(\frac{\theta }{2}\right)}}\\ =PTF\cdot {\epsilon }_{obj}{W}_{\lambda 1-\lambda 2}\left(T_{obj}\right)+PTF\cdot \left(1-{\epsilon }_{obj}\right)\cdot W_{\lambda 1-\lambda 2}\left(T_{amb}\right)\end{array}.$$

(1)

where PTF is the Power-Transfer-Function. The proof for the emitted power which is absorbed by the sensor (the first term) is described in Appendix A.

In the above modeling, we have assumed that the blackbody is larger than the target area (see Figure 5). Under this condition, the PTF depends only on A_D, and the Field of View (FOV). The distance is not a parameter. Furthermore, the PTF varies for different units, but its exact value is determined during the calibration process (see Section 4).

From Expressions (1) it is obvious that there are two unknowns: the object temperature, and emissivity. Hence, two devices are required, with different optical bandpass filters, defined by λ1–λ2 and λ3–λ4:

$$P_{\lambda 1-\lambda 2}\left(\epsilon , T_{obj},T_{amb}\right)=PTF\cdot \epsilon W_{\lambda 1-\lambda 2}\left(T_{obj}\right)+PTF\cdot \left(1-\epsilon \right)\cdot W_{\lambda 1-\lambda 2}\left(T_{amb}\right)$$

(2)

$$P_{\lambda 3-\lambda 4}\left(\epsilon , T_{obj},T_{amb}\right)=PTF\cdot \epsilon W_{\lambda 3-\lambda 4}\left(T_{obj}\right)+PTF\cdot \left(1-\epsilon \right)\cdot W_{\lambda 3-\lambda 4}\left(T_{amb}\right)$$

In this study we are using the following bandpass optical filters, termed narrow and wide, respectively: narrow bandpass λ1–λ2 = 8–14 µm and wide bandpass λ3–λ4 = 5.5–13.5 µm.

2.2.2. With Optics Limiting the Field of View (FOV)

The reflected (second term) of Expressions (1) may introduce significant errors unless the field of view is limited. Figure 6 presents the schematic approach of this study on how to limit the FOV, while Figure 7 exhibits an image of the radiometer.

The edge of the metal tube is placed very close, 1–2 cm, from the forehead or the blackbody during calibration. An appropriate lens placed inside the tube is optional and reduces the effect of distance between sensor and forehead.

3. Modeling and Calibration

3.1. Modeling

The model is based on empirical measurements. We can refer to it as phenomenological or heuristic. We measure the temperature difference between the active pixel to the blind pixel (see Figure 2), multiply it by the gain and then add the offset (see Figure 8):

$$S\left[LSB\right]=gain\cdot \underset{\Delta T_{measrued}}{\underbrace{\left(T_{active}-T_{blind}\right)}}+offset$$

(3)

Figure 8 describes schematically the measured signal as a function of the temperature difference and with two different object emissivity, ε₁ > ε₂.

The offset is measured when there is no temperature difference between the active and blind pixel, ∆T_measured = 0. Ideally, the output signal should be zero. In practice, every sensor, regardless of its specific nature, exhibits offset. The offset depends on the object temperature, emissivity, and ambient temperature. Since the measured plot is very stable in time for a given set of parameters, the offset can be calibrated.

Figure 9 exhibits a typical measured signal, at an ambient temperature of 20 °C as a function of object temperature.

The modeling is based on the observed linearity.

3.2. Calibration

The calibration setup is shown in Figure 10.

The calibration is based on a commercial, extended area blackbody [19], instead of a cavity blackbody. Each channel of the radiometer is calibrated between blackbody temperatures of 32 °C and 42 °C, for at least five different ambient temperatures. Ambient temperatures are determined by the Proportional-to-Absolute-Temperature (PTAT) circuit on the ASIC. Accuracy of ΔT = 0.1 °C and better than Δε = 0.005 is achieved.

The ambient temperature has a strong effect on the results as shown in Figure 11.

The measurements are analyzed according to Appendix C, where the reference value is determined by the blackbody. The accuracy and precision for the calibration are shown in

Table 1. Calibration of accuracy and precision.

	Temperature		Emissivity
Number of Measurements Averaged	Accuracy (Mean)	Precision (Standard Deviation)	Accuracy (Mean)	Precision (Standard Deviation)
1	0.1267	0.103	0.0013	0.001
8	0.0806	0.0531	7.43 × 10−4	5.11 × 10−4
32	0.0739	0.0413	6.50 × 10−4	3.87 × 10−4
100	0.0704	0.0376	5.99 × 10−4	3.56 × 10−4

The values depend on the average number of measurements.

.

4. Measurements Results

4.1. Calibration Validation

The results and the modeling are validated with the same blackbody, where we now measure at arbitrary testing points not included in the calibration. The accuracy and precision for the validation of calibration are shown in

Table 2. Validation of calibration accuracy and precision.

	Temperature		Emissivity
Number of Measurements Averaged	Accuracy (Mean)	Precision (Standard Deviation)	Accuracy (Mean)	Precision (Standard Deviation)
1	0.1218	0.0935	0.0012	9.49 × 10−4
8	0.0725	0.051	6.75 × 10−4	5.01 × 10−4
32	0.0666	0.0351	5.87 × 10−4	3.42 × 10−4
100	0.0664	0.028	5.79 × 10−4	2.68 × 10−4

.

4.2. Forehead Measurements with Different of Cosmetics and Ointments

The emissivity of human skin has acquired considerable importance because of the increasing use of fever screening for elevated temperatures associated with COVID-19 and other illnesses or to consider the effect of applied cosmetics or ointments [20,21].

We applied common creams, cosmetics, lotions, and ointments on the forehead, and then compared the measured surface temperature with the commercial non-imaging no-contact IR thermometers described in Section 5. The applied materials included common make up, moisturizing creams, and skin treatments with vitamin C serum, collagen-based serum and retinol-based serum. As expected, the Digital TMOS radiometer was more sensitive to the changes in surface temperature due to the change in emissivity. The Digital TMOS radiometer exhibited a reduced surface temperature, up to 2 °C, for healthy people and certain cosmetics. Cosmetics and topical medication, which increase the water content in the forehead, increase the emissivity. Since we measured the emitted power by the surface, the increase in emissivity resulted in a decreased surface temperature. Small changes in the emissivity, of the order of several tenth of percentage, already introduced changes in the surface temperature. Unless emissivity is taken into consideration, the IR non-contact temperature measurement can be unreliable and mask people with fevers. This aspect of our study will be reported elsewhere, once a large measurement of the data analyzed with statistical techniques is obtained.

4.3. Correlation between Forehead Measurements and Core Body Temperature—A Disclaimer

The public is familiar with body thermometers (such as mercury thermometers). Hence, all commercial IR remote temperature sensors introduce an offset (a correlation between body and surface temperature). The radiometer reported here is remarkable since it models the offset by using the emissivity in addition to the object’s and the ambient temperatures. We do not present here the offset correction tailored to the Digital TMOS since a large measurement of the data analyzed with statistical techniques is still required.

5. Comparison to Other State-of-the-Art Commercially Available IR Remote Temperature Thermometers

Table 3. Performance comparison of the radiometer understudy and commercial products that are considered state-of-the-art. The superscripts in the table are elaborated in the tables below for each product (referred to as the Notes table). The “channel” refers to the optical bandpass.

Parameter	This Work: Wide Channel [18]	Commercial Medical Grade [22]	Commercial, No Touch Forehead Thermometer [23]	Commercial Gun Thermometer [24]
Notes Table	See Table 4	See Table 5	See Table 6	See Table 7
Concept	2 channels a	1 channel a	1 channel a	1 channel
Packaged Device (mm3)	3.2 × 4.2 × 1.45 b	3 × 3 × 1	Gun thermometer	Gun thermometer
Factory Calibrated	Yes	Yes	Yes	Yes
Field of View (°)	80 c	50	N/A	5
Temp. Sensitivity	2000 (LSB/K)	NETD b = 50 mK	N/A	N/A
Noise (peak to peak)	120 (LSB)]	N/A	N/A	N/A
Resolution (°C)	0.02	0.01	0.1	0.1
Optical Filter (µm)	5.5–13.5	2–14	N/A	8–14
Accuracy (°C)	0.07 d	0.2 c	0.2	±2
Precision (°C)	0.04 d	N/A	N/A	±0.5
Supply Voltage (V)	1.7–3.6	3.3	2 AA battery	9
Supply Current	10 uA	1 mA d	N/A	N/A
Total Power Consumption	0.036 mWatt	3.3 mWatt	N/A	N/A
Ambient temp. range (°C)	15–30 e	15–40 c	15–40 b	N/A
Object temp. range (°C)	35–42 e	35–42 c	35–42 b	−32~
Storage temperature (°C)	−40–85	−20–85	−25–60	N/A
Response Time	85–125 ms	16 ms–2 s	<2 s	500 ms
Total Power Consumption	0.036 mWatt	3.3 mWatt	N/A	N/A

a–e: Notes for the radiometer understudy (see Figure 3 and Figure 7).

summarizes the proposed microsystem performance and compares it with commercial state-of-the-art non-imaging, no-contact IR thermometers. The values are taken from the datasheets of the commercial products. Because of the digital nature of the measurements and the lack of uniform terminology, we follow the terms defined in the Appendix C.

The product is shown in Figure 12:

The product is shown in Figure 13:

The product is shown in Figure 14:

6. Summary and Conclusions

Forehead temperature measurement using an infrared thermometer may be used to safely and rapidly screen for fever to identify victims of COVID-19. Unless both emissivity and temperature are measured, the screening may fail [3,25]. The temperature of the forehead is known to be highly correlated with the internal body temperature and it is easier to access the forehead temperature than the tympanic (ear) temperature [8].

A two-channel radiometer, based on the new nano-metric CMOS-SOI-NEMS Technology, dubbed Digital TMOS [18], enables remote temperature sensing as well as the emissivity sensing of forehead and body temperature of people, with high sensitivity, accuracy, and high resolution.

Currently it is intended for indoor application and for living human bodies with temperatures in the range of 32–42 °C and forehead emissivity around 0.98. We have emphasized the role of emissivity in determining the forehead surface temperature. We have shown that an accuracy as well as a precision better than 0.005 in the emissivity is required to achieve an accuracy of 0.1 °C in the forehead surface temperature.

The Digital TMOS radiometer outperforms commercial gun thermometers or the medical grade sensors, which are based on one IR non-contact sensor. The commercial instruments assume for the user a nominal emissivity, whereas the Digital TMOS radiometer measures both emissivity and temperature with the required fidelity.

We have shown that common cosmetics as well as the measuring lab ambient temperature and body exercise may affect the forehead surface emissivity. Since the accuracy with which we can determine the temperature is strongly dependent on the accuracy with which we can measure the emissivity of the surface, a two-channel radiometer is essential.

Non-contact temperature measurement, known also as “dual channels” or “two-color” IR thermography, has been applied for many years. The different non-imaging commercial instruments use different thermal sensors such as pyroelectric, thermoelectric (thermopiles), and resistive sensors [26]. The most recent development in this family of thermal sensors is the Digital TMOS based on CMOS micro- or nano-machined transistor, vacuum packaged, and combined with CMOS ASIC, where ambient temperature measurement and digital signal processing are performed. This new technology with advanced processor technology increases the system’s stability, reliability, resolution, and speed.

The authors have measured and compared the performance of the Digital TMOS radiometer with state-of-the-art commercial non-imaging IR non-contact thermometers. The measured results exhibit the potential of the Digital TMOS to outperform the existing products.

Forehead temperature measurement using an infrared thermometer may be used to rapidly and safely screen for fever to identify victims of COVID-19. However, a large measurement of the data analyzed with statistical techniques is still required. A clinical study in hospitals is planned to meet this requirement.

Furthermore, this study compares commercial IR thermometers in stable, resting, laboratory conditions. We intend to extend the Digital TMOS radiometer to outdoor measurements.

The role of calibration in any MEMS sensing system is well-known. Calibration determines not only the accuracy but also the cost of commercial products. The role of emissivity in remote temperature measurements is also well established. Detailed information critical to these topics are typically kept private by manufacturers. The academic work reported here focuses on the above topics, hoping to inspire additional studies in the open literature.