Transmission Analysis in Human Body Communication for Head-Mounted Wearable Devices

: As society ages, wireless body area networks (WBANs) are expected to increasingly improve the quality of life of the elderly and disabled. One promising WBAN technology is human body communication (HBC), which utilizes part of the human body as a transmission medium. Communication between head-mounted wearable devices, such as hearing aids, is a potential HBC application. To clarify the HBC transmission mechanism between head-mounted wearable devices, this study analyzes the input impedance characteristics of the transceiver electrodes, transmission characteristics, and electric ﬁeld distributions around and through a detailed head model. The investigation was performed via an electromagnetic ﬁeld simulation. The signal frequency had less effect on the transmission characteristics and electric ﬁeld distributions at 10, 20, and 30 MHz. However, the transmission mechanism between the head-mounted wearable devices was inﬂuenced by the number of electrodes in the transceiver. Moreover, the transmission characteristics between two-electrode transceivers were improved by impedance matching. Finally, the availability of the proposed system was evaluated from power consumption and human safety perspectives.


Introduction
Driven by recent advances in high-density integration technology, information communication devices have evolved from "mobile" to "wearable". The next-generation multiple wearable devices are expected to be connectable through a wireless body area network (WBAN), which is defined as a network of devices in close proximity to a person's body [1]. In 2012, the IEEE 802.15.6 working group standardized WBAN [2]. WBANs are primarily applied in medical and welfare devices, which must be connected for the exchange of data such as vital signs data. As society ages, the role of these technologies is expected to increase for improving the quality of life of the elderly and disabled. Many of these devices adopt short-range wireless communication standards (within a few meters around the body). Examples of the commonly used standards are Bluetooth Low Energy and ZigBee (both operate at 2.4 GHz). In Japan, an extremely low-power radio system, which operates in the 315 MHz band, and a specified low-power radio system, which operates in the 400 MHz band [3,4], are also popular. Usually, the human body is regarded as an obstacle in wireless communication because the human body absorbs high-frequency electromagnetic waves in the 315 MHz to 2.4 GHz range [5,6].
Human body communication (HBC), which utilizes a part of the human body as a transmission medium, has attracted considerable attention as one of the methods for WBAN data transmission [7]. In HBC, a radio-frequency signal flowing through the body and in the near-field electric field or in a surrounding quasi-electrostatic field contributes to the communication among devices [8]. Meanwhile, the human body is a conductive

Simulation Models
In this study, we assumed signal transmission between two wearable transceivers worn on each side of the head. For comparison, we also analyzed conventional wireless communications at 2.45 GHz. Analyses were performed using the simulation software XFdtd (Remcom Inc., State College, PA, USA), which employs the FDTD method.

Human Body Models
To evaluate the trade-off between calculation accuracy and time, we simulated three types of human body models with different simplifications. Figure 2 shows the whole-body human model "TARO" provided by the National Institute of Information and Communications Technology [48]. Panel (a) of this figure shows the model dimensions and a cross-sectional view in the x-z plane, and (b) shows the positions of the transceiver antennas in the 2.45-GHz WC simulation. This model is useful for calculating signal propagations because it comprises 51 different biological tissues (including skin, fat, and muscle), each with its own electrical properties. In this study, the electrical property of each biological tissue was assumed to be frequency-dependent [9].

Simulation Models
In this study, we assumed signal transmission between two wearable transceivers worn on each side of the head. For comparison, we also analyzed conventional wireless communications at 2.45 GHz. Analyses were performed using the simulation software XFdtd (Remcom Inc., State College, PA, USA), which employs the FDTD method.

Human Body Models
To evaluate the trade-off between calculation accuracy and time, we simulated three types of human body models with different simplifications. Figure 2 shows the whole-body human model "TARO" provided by the National Institute of Information and Communications Technology [48]. Panel (a) of this figure shows the model dimensions and a cross-sectional view in the x-z plane, and (b) shows the positions of the transceiver antennas in the 2.45-GHz WC simulation. This model is useful for calculating signal propagations because it comprises 51 different biological tissues (including skin, fat, and muscle), each with its own electrical properties. In this study, the electrical property of each biological tissue was assumed to be frequency-dependent [9]. Figure 3a shows the detailed human head model extracted from the whole-body model "TARO". This head model, which consists of 23 biological tissues, reduces the memory requirements, CPU/GPU power, and calculation time because the trunk is removed. The head model in Figure 3b is further simplified by assuming homogenous muscle tissue. The homogenous muscle model is popular because it is the simplest model for the qualitative analysis of WBAN, including HBC [22,49]. In addition, we compared the results of a homogenous muscle arm model and a three-layered arm model composed of skin, fat, and muscle. The patterns of electric field distribution were similar in both models, but some quantitative values differed between the models [50]. To reduce the edge effect of the electric field, the front and back surfaces of the model were curved into a shape resembling the actual head shape. Both temporal regions of the model were flattened for Electronics 2021, 10, 1213 4 of 15 perfect attachment of the electrodes. The model dimensions were based on the average body shape of a Japanese adult [51]. The model composed only of muscle tissues was also given a frequency-dependent electrical property.  Figure 3a shows the detailed human head model extracted from the whole-body model "TARO". This head model, which consists of 23 biological tissues, reduces the memory requirements, CPU/GPU power, and calculation time because the trunk is removed. The head model in Figure 3b is further simplified by assuming homogenous muscle tissue. The homogenous muscle model is popular because it is the simplest model for the qualitative analysis of WBAN, including HBC [22,49]. In addition, we compared the results of a homogenous muscle arm model and a three-layered arm model composed of skin, fat, and muscle. The patterns of electric field distribution were similar in both models, but some quantitative values differed between the models [50]. To reduce the edge effect of the electric field, the front and back surfaces of the model were curved into a shape resembling the actual head shape. Both temporal regions of the model were flattened for perfect attachment of the electrodes. The model dimensions were based on the average body shape of a Japanese adult [51]. The model composed only of muscle tissues was also given a frequency-dependent electrical property.   Figure 4 shows the structures of the transceivers (transmitter: TX, receiver: RX) in the electrode model of the HBC simulation. One transceiver was located at each temple. In this figure, panel (a) presents the two-electrode model and panel (b) presents single-electrode models. The two-electrode model comprises a circuit board, a signal electrode, a ground electrode, wires, and a feeding point (transmitter) with an output impedance of 50 Ω or a passive load (receiver) with a 50 Ω resistance. We adopted the transmitter and receiver designed in our previous studies (a = 8, b = 24, d = 8, L1 = 24, L2 =  The two-electrode model comprises a circuit board, a signal electrode, a ground electrode, wires, and a feeding point (transmitter) with an output impedance of 50 Ω or a passive load (receiver) with a 50 Ω resistance. We adopted the transmitter and receiver designed in our previous studies (a = 8, b = 24, d = 8, L 1 = 24, L 2 = 24, and h = 10 mm). The electrodes of these devices were designed to provide input and output impedances of 50 Ω in a muscle homogenous model [52]. The single-electrode model consisted of a circuit board, a signal electrode, wires, and a feeding point (transmitter) with an output impedance of 50 Ω or a passive load (receiver) with a 9 kΩ resistance. All transceivers were composed of a perfect electric conductor (PEC) material. The undersurfaces of the electrodes were designed for firm attachment to the skin when applied to the detailed head model.  As the transceiver antenna of the 2.45 GHz WC, we employed a dipole antenna adjusted to resonate at 2.45 GHz in free space. The antenna was constructed from a PECwire element with a length of 54 mm. A feeding point (transmitter) with 50 Ω output impedance or a 50 Ω resistance passive load (receiver) was placed at the center of the antenna element.

Simulation Conditions
The computing space of the FDTD simulation was represented by non-uniform grids. As the grids became more distant from the vicinity of the transceiver electrodes and antennas, their size gradually increased. The grid size was 1 mm around the transceiver electrodes and 5 mm at the edge of the computing space. The padding, or free-space distance between the model edge and absorbing boundary, was 20 cells thick. The absorbing boundary was a perfectly matched layer composed of seven sublayers in the two-electrode and 2.45 GHz antenna transceiver models, and 15 layers in the single-electrode transceiver model. A lumped port was selected as both the feeding point and receiving load of the transceivers. A broadband pulse was fed at the feeding point. The time steps of the HBC and 2.45 GHz WC calculations were set to 1.872 and 1.048 ps, As the transceiver antenna of the 2.45 GHz WC, we employed a dipole antenna adjusted to resonate at 2.45 GHz in free space. The antenna was constructed from a PECwire element with a length of 54 mm. A feeding point (transmitter) with 50 Ω output impedance or a 50 Ω resistance passive load (receiver) was placed at the center of the antenna element.

Simulation Conditions
The computing space of the FDTD simulation was represented by non-uniform grids. As the grids became more distant from the vicinity of the transceiver electrodes and antennas, their size gradually increased. The grid size was 1 mm around the transceiver electrodes and 5 mm at the edge of the computing space. The padding, or free-space distance between the model edge and absorbing boundary, was 20 cells thick. The absorbing boundary was a perfectly matched layer composed of seven sublayers in the two-electrode and 2.45 GHz antenna transceiver models, and 15 layers in the single-electrode transceiver model. A lumped port was selected as both the feeding point and receiving load of the transceivers. A broadband pulse was fed at the feeding point. The time steps of the HBC and 2.45 GHz WC calculations were set to 1.872 and 1.048 ps, respectively.

Effects of Human Body on Signal Transmission
This subsection examines the effects of the human body on communications between the head-mounted wearable devices in the whole-body human model ( Figure 2). The 2.45 GHz WC and HBC were analyzed in two environments: one with the human body (LOS) and the other without the human body (NLOS). In the 2.45 GHz WC simulation, the impedances of the feeding point and receiving load of the transceiver antennas were both set to 50 Ω. In the HBC simulation, the electrode configuration was a two-electrode transceiver, and the carrier frequency was 10 MHz. The impedances of the feeding point and receiving load of the transceiver electrodes were both set to 50 Ω.
Panels (a)-(d) of Figure 5 show the electric field distributions calculated under each condition, and Table 1(a)-(d) lists the transmission characteristics S 21 between the transceivers under the conditions of Figure 5. The electric field distribution was observed in the x-z plane, which included the feeding point, as shown in Figure 3. The dotted curve in Figure 5 outlines the boundary between the human body model and free space. As shown in Figure 5a,b, the electric field was blocked by the human body between the transceiver antennas, and the signal in the 2.45 GHz WC was not propagated. The human body deteriorated the S 21 by 50 dB or more (Table 1(a),(b)). From these results, we inferred that the 2.45 GHz WC is unsuitable in the NLOS environment caused by the human body such as the communication between head-mounted wearable devices. In contrast, the electric field was propagated through the head in the HBC scenario (Figure 5c,d), and the head tissues improved the S 21 by 46 dB (Table 1(c),(d)). On the contrary, the electric field was attenuated by > 50 dB in the free space at 10 cm from the feeding point. As a result of this exponential decrease in the electric field, the signal propagated along the body without interfering with other devices. This feature is beneficial for confidential communication and EMC. In the NLOS environment, the S 21 was approximately 10 dB larger in the HBC scenario than in the 2.45 GHz WC scenario. Furthermore, as shown in Section 4.4, S 21 can be further improved by impedance matching. These results demonstrate that the proposed system achieves stable communication and that HBC reduces the power consumption.

Simplification of Human Body Model
To properly meet the trade-off between the computer resources and the accuracy of the calculation results, the model should be simplified but without significantly affecting the calculated transmission characteristics and electric field distributions. The model shown in Figure 3a is the detailed human head model extracted from the whole-body model. Since it removes the other parts, this model significantly conserves the computer resources. However, as the trunk and limbs may affect the distribution and transmission characteristics of the electromagnetic field, this model must be confirmed. As mentioned earlier, the model in Figure 3b is further simplified to a simple solid with homogenous muscle, which conserves more computer resources than the detailed head model. However, the different electrical properties of the different biological tissues may greatly influence the transmission parameters of HBC. Therefore, the homogenous head model without tissue structure must also be validated. In the present study, the simplified models were validated using a two-electrode transceiver and a 10-MHz carrier frequency. The impedance of the feeding point and receiving load of the transceiver electrodes were both set to 50 Ω.     Figure 6 shows the electric field distribution calculated under each condition. Again, the electric field distribution was observed on the x-z plane, which includes the feeding point, and the white dotted curve delineates the boundary between the human body model and free space. Table 2 lists the resistive and imaginary components, Re(Z in ) and Im(Z in ), respectively, of the input impedances of the electrodes. It also lists the transmission characteristics S 21 between the transceivers. Comparing panels (a) and (b) of Figure 6, we observe that the electric field distributions around and inside the head were almost identical in the whole-body and detailed head models, for the following reasons. At a carrier frequency of 10 MHz, the wavelength is 30 m, which is sufficiently larger than the total length of the human body, and the electric field distribution was determined by the local conditions i.e., the head. Furthermore, as shown in Table 2(a),(b), all transmission parameters were almost unaffected by the body parts below the neck.
In contrast, comparing Figure 6b,c, the electric field distribution was significantly altered in the head model with simplified shape and tissue structure. Z in was also changed significantly, and S 21 increased by more than 1 dB (Table 2(b),(c)). As evidenced in the different electric field distributions, the signal transmission in HBC was enabled by the complicated distribution of biological tissues with their different electrical properties. A similar investigation performed at 20 and 30 MHz confirmed that when analyzing the transmission parameters of HBC, the body parts below the neck could be ignored, but the tissue structure was important. Overall, the detailed head model shown in Figure 3a was considered as the best model considering the trade-off between the resource demands and accuracy of the computational results. Therefore, the detailed head model was used in subsequent studies.

Transceiver Electrodes Configuration and Carrier Frequency in HBC
In communications between wearable devices worn on the body other than the head, a two-or single-electrode transceiver is employed to improve the transmission efficiency in a given situation. However, the transmission mechanism can differ when wearable devices are worn on both sides of the head from that when worn on the arm because the transmitter and receiver are at separate locations on the human body. The two-electrode and single-electrode models, as shown in Figure 4a,b, respectively, have been proposed as transmitter structures. Considering their HBC-coupling mechanisms, the two-electrode and single-electrode models are classified as galvanic and capacitive coupling HBC models, respectively. In this subsection, we clarify the transmission mechanisms for different numbers of contact electrodes, input impedance characteristics of the transceiver electrodes Z in , transmission characteristics S 21 , and electric field distributions around/inside the detailed head model. Moreover, the carrier frequency was varied as 10, 20, and 30 MHz focusing on the physical layer defined in ISM band, IEEE 802.15.6.

Two-Electrode Transceiver
Figure 7a-c shows the electric field distributions around and inside the human head in the x-z plane, including the feeding point of the transceiver. The Re (Z in ) and Im (Z in ) impedance components decreased with increasing carrier frequency (Table 3), and S 21 ranged from −64.3 to −64.8 dB. Signal frequencies of 10, 20, and 30 MHz negligibly affected the transmission characteristics and electric field distributions, implying that the transmission characteristics are stable in the 10-30 MHz range. Therefore, we can select the carrier frequency that best fits the available frequency bands, avoids interference from other wireless systems, and satisfies the hardware restriction. Furthermore, the electric field generated by the two-electrode transceiver penetrated the head interior because the high-frequency current was concentrated between the signal and ground electrodes. The penetrating electric field had a significant effect on the transmission. This transmission mechanism of a two-electrode transceiver is quite different from that of 2.45 GHz WC. The improved transmission characteristics of our system are promising for future development; however, the system must first be evaluated from a human safety viewpoint.    impedance components decreased with increasing carrier frequency (Table 3), and S21 ranged from −64.3 to −64.8 dB. Signal frequencies of 10, 20, and 30 MHz negligibly affected the transmission characteristics and electric field distributions, implying that the transmission characteristics are stable in the 10-30 MHz range. Therefore, we can select the carrier frequency that best fits the available frequency bands, avoids interference from other wireless systems, and satisfies the hardware restriction. Furthermore, the electric field generated by the two-electrode transceiver penetrated the head interior because the high-frequency current was concentrated between the signal and ground electrodes. The penetrating electric field had a significant effect on the transmission. This transmission mechanism of a two-electrode transceiver is quite different from that of 2.45 GHz WC. The improved transmission characteristics of our system are promising for future development; however, the system must first be evaluated from a human safety viewpoint.

Single-Electrode Transceiver
Figure 8a-c show the electric field distributions around and inside the human head in the x-z plane, including the feeding point of the transceiver. As shown in Table 4, the signal frequencies at 10, 20, and 30 MHz had a negligible effect on the S 21 and electric field distributions but a significant effect on the Im(Z in ) because the imaginary impedance includes the capacitive reactance (determined by the floating capacitance between the circuit board and the human body), which depends on the carrier frequency. The electric field of the single-electrode transceiver moved around the head without penetrating it (Figure 8). This transmission mechanism was similar to that of 2.45 GHz WC, implying that the electric current in the biological tissues was small. Although this finding is favorable from a human safety viewpoint, the transmission characteristics were degraded.
distributions but a significant effect on the Im(Zin) because the imaginary impedance includes the capacitive reactance (determined by the floating capacitance between the circuit board and the human body), which depends on the carrier frequency. The electric field of the single-electrode transceiver moved around the head without penetrating it (Figure 8). This transmission mechanism was similar to that of 2.45 GHz WC, implying that the electric current in the biological tissues was small. Although this finding is favorable from a human safety viewpoint, the transmission characteristics were degraded (a) (b) (c)

S21 Improvement by Impedance Matching and Exposure Assessment for Human Safety
In Section 4.3, it was suggested that the transmission characteristics of the two-electrode structure are appropriate for a head-mounted wearable system. To further stabilize the communication and lower the power consumption, the transmission characteristics can be improved by impedance matching. The reactance components of Zin were removed by inserting an inductor in series with the resistance of the feeding point and the receiving load. Table 5 lists the values of the circuit elements (resistance and inductance) of the transceivers at each frequency. These values were calculated to match the impedance of the transceiver circuit with Zin given in Table 3. Table 5 also lists the calculated transmission characteristics S21 under the impedance-matched conditions. Comparing Tables 3 and 5, we find that the impedance matching improved the S21 by 2.2-  In Section 4.3, it was suggested that the transmission characteristics of the twoelectrode structure are appropriate for a head-mounted wearable system. To further stabilize the communication and lower the power consumption, the transmission characteristics can be improved by impedance matching. The reactance components of Z in were removed by inserting an inductor in series with the resistance of the feeding point and the receiving load. Table 5 lists the values of the circuit elements (resistance and inductance) of the transceivers at each frequency. These values were calculated to match the impedance of the transceiver circuit with Z in given in Table 3. Table 5 also lists the calculated transmission characteristics S 21 under the impedance-matched conditions. Comparing Tables 3 and 5, we find that the impedance matching improved the S 21 by 2.2-4.3 dB at each frequency. That is, the electrodes matched the impedance not only to the human body but also between the body and the surrounding space. This result confirms that impedance matching effectively improved the transmission characteristics in the proposed system. Finally, the human safety of the proposed system was assessed by the specific absorption rate (SAR), which determines the human exposure to an electromagnetic field. A safety assessment is necessary because the electric field is distributed inside the head. The SAR (in W/kg) defines the power of a radio-frequency electromagnetic field absorbed by one kilogram of human tissue. It is calculated from the electric field within the tissue as SAR = (σE 2 /ρ), where E is the RMS electric field, σ is the electrical conductivity of the tissue sample, and ρ is the sample density. In the averaging scheme of SAR, 10 g averaging was selected because the temperature increase in the head is reasonably correlated with SAR in this averaging scheme [53]. In the SAR calculation, we used S 21 = −61.7 dB at 10 MHz because this value was smallest among those of the studied frequencies. The ultimate sensitivity of a conventional wireless communication receiver (such as a Bluetooth receiver) is approximately −70 dBm [54]. In the transmitter/receiver configuration of this study, the feeding point required 150 µW (−8.2 dBm) to provide −70 dBm at the receiver. Under this condition, the peak SAR per 10 g of head tissue was 0.014 W/kg between the transmitter electrodes. This value is below the regional absorption limit (2 W/kg per 10 g of any tissue) in the Safety Guidelines for Use of Radio Waves established by the Ministry of Internal Affairs and Communications [55]. Furthermore, Figure 9 shows the SAR distribution inside the human head in the x-z plane, including the feeding point of the transceiver. The peak SAR (0.014 W/kg) was normalized to 0 dB. The white parts in the distribution are cavities filled with air, in which the SAR is not defined. The SAR peaked directly below the transmitter electrodes and was attenuated with increasing distance from the transmitter. For example, in brain tissue, the SAR was 30 dB-40 dB lower than the peak value. These results clarify that an impedance-matched two-electrode structure transmitter/receiver system achieves sufficient transmission characteristics and satisfies the human safety requirements. one kilogram of human tissue. It is calculated from the electric field within the tissue as SAR = (σE 2 /), where E is the RMS electric field, σ is the electrical conductivity of the tissue sample, and  is the sample density. In the averaging scheme of SAR, 10 g averaging was selected because the temperature increase in the head is reasonably correlated with SAR in this averaging scheme [53]. In the SAR calculation, we used S21 = −61.7 dB at 10 MHz because this value was smallest among those of the studied frequencies. The ultimate sensitivity of a conventional wireless communication receiver (such as a Bluetooth receiver) is approximately −70 dBm [54]. In the transmitter/receiver configuration of this study, the feeding point required 150 μW (−8.2 dBm) to provide −70 dBm at the receiver. Under this condition, the peak SAR per 10 g of head tissue was 0.014 W/kg between the transmitter electrodes. This value is below the regional absorption limit (2 W/kg per 10 g of any tissue) in the Safety Guidelines for Use of Radio Waves established by the Ministry of Internal Affairs and Communications [55]. Furthermore, Figure 9 shows the SAR distribution inside the human head in the x-z plane, including the feeding point of the transceiver. The peak SAR (0.014 W/kg) was normalized to 0 dB. The white parts in the distribution are cavities filled with air, in which the SAR is not defined. The SAR peaked directly below the transmitter electrodes and was attenuated with increasing distance from the transmitter. For example, in brain tissue, the SAR was 30 dB-40 dB lower than the peak value. These results clarify that an impedance-matched two-electrode structure transmitter/receiver system achieves sufficient transmission characteristics and satisfies the human safety requirements.

Conclusions
In an electromagnetic field simulation, this study clarified the HBC transmission mechanism between a pair of head-mounted wearable devices, the input impedance characteristics of the transceiver electrodes, the transmission characteristics, and electric field distributions around a detailed head model. Analyses using the whole-body human model revealed that HBC is more suitable than 2.45 GHz WC for the NLOS environment caused by the human body. In addition, the availability of the detailed head model was confirmed in terms of the trade-off between the computer resources and accuracy of the calculation results. The signal frequency (10, 20, or 30 MHz) had less effect on the transmission characteristics and electric field distributions of the wearable devices, but the transmission mechanism between the head-mounted wearable devices was influenced by the number of transceiver electrodes. Moreover, the transmission characteristics between the two-electrode transceivers were improved by impedance matching. Finally, the power consumption and human safety analyses confirmed the availability of the proposed system.
In the next step of our research, we will develop a more simplified circuital model based on the transmission mechanism clarified herein because simple models such as the distributed circuital model [21] are useful for the hardware design phase. Moreover, using the obtained transmission parameters, we will investigate physical layer issues such as the electrode structure design and modulation strategies. Comparisons with other HBC systems using low-complexity devices [34] are also important to develop the practical application.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest.