Numerical Analysis of Users’ Body Effects on a Fourteen-Element Dual-Band 5G MIMO Mobile Terminal Antenna

This paper studies the performance of a 14-element dual-band 5G MIMO antenna with body effect and an antenna selection. The antenna operates in the lower frequency band from 3.10 to 3.85 GHz for 5G bands of LTE 42 and 43, whereas the upper band is from 5.60 to 7.20 GHz for the newly assigned 6 GHz WiFi band. The antenna performance was evaluated in free space and under the effect of human body in three scenarios, namely one-hand, two-hands, and call mode. All resulting envelope correlation coefficients are less than 0.3. On the other hand, the free space efficiency of the antenna elements is between 41 % to 77 % in the lower band and 61 % to 95 % in the upper band. Meanwhile, the efficiency of the elements varies depending on the interaction between each element with the hands and the head, with the least obstructed element efficiency degrading to 8 % relative to free space levels. Due to the body effect, multiplexing efficiency loss varies from 2.68 dB to 5.93 dB and the capacity loss is from 14 % to 38 %. Finally, a selection algorithm is used to assess the performance of the overall antenna when 12, 10 and 8 antenna elements are selected for operation. In free space and with number of elements 12, 10 and 8, the system achieves 91.7 %, 80 % and 67 % from the capacity of the 14-element MIMO antenna, respectively. However, in the vicinity of the body, these values increase to 96.3 %, 85.0 % and 72 % from the capacity of the 14-element MIMO antenna, respectively. This signifies that the less-contributing elements as a result of body blockage can be deselected in order to preserve system resources that these elements consume while contributing less significantly to the performance.


I. INTRODUCTION
Multiple input multiple output (MIMO) antennas are deployed on both sides of the wireless communication link The associate editor coordinating the review of this manuscript and approving it for publication was Luyu Zhao . to achieve higher performance over single input single output (SISO) antennas without demanding higher bandwidth or transmit power [1]. Key advantages of MIMO antennas are multiplexing, beamforming, and spatial diversity to increase data rates, improving link reliability and achieve protection against interference [2], [3]. VOLUME 10, 2022 This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ In the 3G and 4G mobile communication systems, the number of MIMO antenna elements on a mobile terminal was between two and four elements. This number of elements is adequate to fulfill the performance requirements of these systems [4]. However, in 5G systems, more antenna elements are required on the mobile terminal to achieve the higher performance envisioned [5]. Nevertheless, the higher number of antenna elements on the mobile terminal requires more circuitry and signal processing [6]. Furthermore, the limited space on the chassis may restrict the number of elements that can be installed. Researchers have proposed 5G MIMO antennas for mobile terminals with different number of elements and different operating frequency bands. For instance, [7]- [10] proposed 8-element MIMO antennas for 5G mobile terminals, a 10-element MIMO antennas in [11]- [13], a 12-element MIMO antenna in [4] and [14], 14-element and 16-element MIMO antennas in [15] and [16], respectively. Even as high as 20-element MIMO antenna was proposed in [17]. However, in practice, the performance of MIMO antennas on mobile terminals is limited by aspects such as the human body's effect. This effect eventually will result in a major degradation in the link-level performance of the wireless system [18]. Therefore, the human body needs to be considered when evaluating the performance of MIMO antennas of mobile, and consequently, techniques to alleviate this effect need to be proposed [19]- [21].
In this paper, a 14-element MIMO antenna for 5G applications is proposed. The performance of the antenna is evaluated in free space and when the device is used in the vicinity of human body in three interaction scenarios -one-hand grip, two-hand grip, and call mode. Different metrics are used to evaluate the antenna performance in the multipath propagation environment.
Finally, an algorithm for antenna selection and removal of elements with least contribution is used to evaluate antenna capacity when the device is used in free space and under the three scenarios of interaction with the human body. This work's contribution is the application of the antenna selection algorithm, in addition to the performance evaluation of the antenna in various interactions scenarios with the user's body. This is done to identify and exclude low-contribution antenna elements caused by body blockage, thereby saving system resources that these elements would otherwise consume.
The organization of the paper is as follows. The next section presents the antenna details and the selected scenarios of interaction with the human body. Then, the radiation performance of the antenna is discussed in the following section in terms of S-parameters, radiation pattern, efficiency and specific absorption rate. Section 4 then presents the envelope correlation coefficient (ECC) results, followed by the multiplexing efficiency and ergodic capacity results in sections 5 and 6, respectively. Section 7 presents the antenna capacity results when the antenna selection algorithm is applied, before the conclusion is presented in section 8. Fig. 1 shows the geometry of the proposed 14-element MIMO antenna integrated on a mobile terminal chassis. All antenna dimensions (length and height) are reasonably selected for practical mobile phones. The proposed antenna consists of a monopole with two branches, and is placed on a metallic ground plane sized at 160 mm × 80 mm × 0.3 mm. The antenna is fed from the first L-shaped arm using a discrete port. Two unequal monopole arm lengths are stacked on top of each other and optimized to obtain a dual band characteristic. The longer monopole branch is designed to operate in the lower 5G frequency band, whereas the shortest branch for operation in the whole 6-GHz WiFi band. The thickness of the metal is same for the whole monopole. The orientation of the antennas is also optimized to get better reflection coefficient and isolation. As mobile terminals are mostly used in the vicinity of human body, the proposed antenna performance is evaluated in the vicinity of the human body in three situations -onehand grip, tow-hand grip and finally in the call mode. Human body phantoms in the CST software are used to evaluate the performance of these three interactions, as shown Fig. 2. Fig. 3 shows the S-parameters of the antennas in operating bands. The lower band (LB) has a frequency bandwidth of 0.75 GHz from 3.10 to 3.85 GHz. This range includes the 5G bands of LTE 42 and 43, from 3.4 to 3.8 GHz [22]. The far-field radiation patterns of the antenna elements in this band are simulated every 50 MHz, resulting in 16 frequency points for every scenario among the four investigated scenarios (one free space and three scenarios with the human body). In addition to that, the upper band (UB) covers a wider frequency bandwidth of 1.6 GHz from 5.60 GHz to 7.20 GHz. This range contains the newly allocated 6-GHz WiFi band, operating from 5.925 to 7.125 GHz. This WiFi band was expanded by the Federal Communications Commission to meet the growing bandwidth demands of WiFi channels [23]. The far-field radiation patterns of the antenna elements are simulated every 100 MHz (with total of 18 frequency points) in each evaluation scenario.

B. FAR-FIELD RADIATION PATTERN
The far-field radiation patterns of four selected antenna elements (elements 2, 6, 10 and 13) in the x-y plane are shown in Fig. 4 at 3.5 GHz (LB). The radiation pattern is simulated in free space and when the device is used in the vicinity of the human body for the three cases. The selected antenna elements are located on each of the four sides of the chassis. Due to this, the effects of the human body for the three on-body cases is noticeable, as the radiation pattern of elements are in direct contact with the hand and/or the head. For instance, elements 2 and 6 are obstructed in the one-hand scenario, thus their radiation patterns are more distorted compared to the other two antenna elements which are further away in this scenario. On the contrary, the patterns for elements 10 and 13 are obstructed by the two-hand grip compared to elements 2 and 6. Finally, the performance of the elements in the call mode is similar to the one-hand case, with more deterioration. However, detailed quantitative results of interaction effect between the human body and each antenna element will be discussed in the next section.
In addition, the 3D realized gain of the same antenna elements which are 2, 6, 10, 13 in free space are shown in Fig. 5.

C. EFFICIENCY
The efficiency of the 14 antenna elements in both bands (in free space and the three on-body cases) is illustrated in Fig. 6. In free space, all elements show similar efficiency in terms of values and increase/decrease behavior with frequency. Another observation is that the efficiency of the antenna elements is higher in the UB than in the LB. The free space efficiency in the LB ranges from 41 % to 77 %, whereas the performance is higher in the UB with values from 61 % up to 95 %. On the contrary, the efficiency of each antenna element in the vicinity of the human body varies depending on the interaction between the element and the hands and/or the head. When the device is held by one hand, the elements on the long sides of the chassis are more affected by the hand, whereas the elements on the short sides of the chassis remained free. Therefore, efficiencies of obstructed elements degraded severely, whereas the unobstructed elements maintained higher efficiencies. In the LB, the best performing antenna is element 9 with an efficiency ranging from 33 % to 55 %. On the other hand, the most affected antenna is element number 6, which efficiency dropped to values between 10 % and 28 %. Meanwhile, in the UB, elements 5 and 9 preserved VOLUME 10, 2022 their performance, producing the highest efficiency between 55 % and 74 %, whereas element 3 showed the low efficiency from 14 % to 27 %.
An important observation is on the higher efficiency of element 11, which is located on the top side of the chassis. The index finger placed behind this element caused degraded its efficiency to be the smallest among elements on the short sides of the chassis. The efficiency of element 11 is from 19 % to 34 % in the LB and around 25 % in the UB.
Next, the antenna interaction when the device is held by the two hands is studied. Generally, in this case the elements located on the short sides of the chassis are covered by 2086 VOLUME 10, 2022 hands, whereas elements on the long sides are free. Thus, the efficiency behavior of elements when the chassis is held using the two-hand grip is the opposite to the performance of the one-hand grip. With the former, the highest efficiency in the LB is seen on elements 2, 3, 6, 7, and the efficiency of these elements increases with frequency from 30 % to 64 %. On the contrary, the most affected elements in the LB with the two-hand grip are 9 and 12, with severe efficiency degradation ranging from 12 % to 29%. Furthermore, the efficiency of elements under the two-hand grip on the UB behaves similarly with the LB in this interaction scenario. Elements 1, 2, 3, 6, 7 sustained the highest efficiencies, from 60 % to 78 %, while the least efficiency is produced by the obstructed elements 9 and 12 with efficiency between 17 % and 20 %.
Finally, the third interaction scenario with the body is when the device is evaluated in the call mode. Due to the direct contact between the antenna elements and the larger body tissue mass from hand and head, the resulting performance in this case is the worst. Similar to the one-hand scenario performance, the elements on the long sides of the chassis deteriorated more severely compared to the elements on the short sides of the chassis. In the LB, all elements exhibited similar efficiency, with a maximum efficiency of 30 % shown by element 9, and the worst efficiency of 8% by element 12. Despite not being obstructed by the hand, the low performance of element 12 is caused by the direct contact with the head. On the other hand, the efficiency in the UB is up to 51 % produced by element 9, and is as low as 10 % seen by element 6.
The results from this section showed that up to 14 antenna elements can be placed on the chassis with satisfactory performance in free space. However, the interaction with human hands and head is unavoidable and resulted in the efficiency degradation, depending on the scenario of interaction. This fact must be considered to ensure that the high number of elements will improve the performance of MIMO antennas in practice. One of the methods alleviate the effects of blockage is by applying antenna selection algorithms aimed at reducing resources consumed by the obstructed elements and maximizing the contribution of the unobstructed elements to optimize the overall system performance.

D. SPECIFIC ABSORPTION RATE (SAR)
The Specific Absorption Rate (SAR) is the amount of electromagnetic energy absorbed by a human body and is measured in watts per kilogram (W/kg). To ensure that a wireless communications device meets the requirements for radio frequency (RF) exposure, wireless devices with radiating parts that are close to the human body must undergo SAR testing [24]. In this work, we show SAR analysis for only call mode where the mobile phone is close to the human head and if the SAR exceeds the maximum limit inside the head, it causes health hazards to the human brain. The maximum SAR over 10g of tissue is 1.92 W/kg. This value is below the maximum allowed value of 2.0 W/kg over 10g [25]. SAR results show that the proposed antenna is safe to use in its call mode. The SAR values at 3.5 GHz and 6 GHz are shown in Fig. 7 at frequencies of 3.5 GHz and 6 GHz.

IV. ENVELOPE CORRELATION COEFFICIENT
Envelope correlation coefficient (ECC) between two antennas characterizes how independent are their far-field radiation patterns. ECC is calculated from the relation given in [26]. Ideally, two independent antennas will have null ECC, whereas a set of fully correlated antennas have ECC of unity. For an acceptable MIMO antenna performance in practical systems, ECC value should be less than 0.5 [27] and this value was revised to be below 0.3 in 4G systems [28]. However, the work in [29] showed that ECC is highly dependent the distribution of the propagation environment. More specifically, the ECC values below these two thresholds in the uniform environment can be higher when narrow beam incident waves are considered. Assuming a uniform environment, ECC is evaluated as follows [26]: where E i (θ, φ) denotes the i th antenna's complex threedimensional (3-D) radiated far-field pattern. The number of ECC values of an n-element MIMO antenna is n(n − 1)/2 at every frequency point (simulated or measured). Thus, for the 14-port MIMO antenna proposed in his VOLUME 10, 2022 work, there will be 91 values at every simulation frequency point. For brevity reasons, ECC curves are omitted from this work. Instead, the maximum ECC values in the lower and higher bands in free space and with the different interaction scenarios with the human body will be reported. Table 1 summarizes these maximum ECC values assuming a uniformly distributed incident signal. In both frequency bands the ECC is below the 0.3 threshold. Furthermore, ECC values in the LB are higher than in UB, due to LB's longer wavelengths with the same physical distance. Hence, the separation distance in terms of wavelength between antenna elements become shorter in the LB. On the other hand, ECC increases when the device is used in the vicinity of the human body, with the highest ECC produced when the device is used in call mode in both bands. These ECC values indicate that up to 14 antenna elements can be integrated onto a common mobile terminal to operate in both frequency bands with acceptable ECC performance.

V. MULTIPLEXING EFFICIENCY
Multiplexing efficiency (η mux ) is a parameter related to the signal to noise power ratio (SNR) of MIMO antennas on 2088 VOLUME 10, 2022 mobile terminals. It is defined as the quotient of the SNR of the MIMO antenna under test (SNR AUT ) to the SNR of an ideal MIMO antenna system (SNR IID ). The ideal SNR is obtained from an independent and identically distributed (IID) MIMO antenna. Both of antennas are assumed to have the same number of elements, and they achieve the same ergodic capacity at these SNRs [30]. Fig. 8 illustrates the concept of the multiplexing efficiency. Mathematically, η mux is defined as: or when both SNRs are in decibels (dB): η mux can be calculated directly from the definition. However, this direct approach is computationally complex, and does not show a direct link between η mux and the design metrics of the MIMO antenna (such as ECC and efficiency). Therefore, an approximate, closed-form η mux formula was derived in [31] and used to evaluate the performance of MIMO mobile terminal antennas in [32]- [34], as follows: where η i is the efficiency of the i th antenna element and det(R) is the determinant of the complex correlation coefficient matrixR, which is constructed as follows [35]: where ρ c,ij is the complex correlation coefficient between ports i and j, and * is the conjugate operator. Besides the low computation complexity of η mux from this closed-form equation, it consists of two main parts; a first part is M M i=1 η i which the geometrical mean of individual efficiencies of VOLUME 10, 2022 the antenna elements. The second part is M det(R) which depends on the correlation between antenna elements. This term gets close to 0 dB when all complex correlation coefficient between the AEs are close to 0. Optimizing these two parts enhances η mux of the MIMO antenna. Fig. 9 presents the multiplexing efficiency of the 14-element MIMO antenna in free space and in the three interaction scenarios with the human body. It observed that the contribution of the M det(R) term is insignificant in the deterioration of η mux compared to the contribution from efficiency term. This is especially evident in the UB, where the ECC values are very low and resulting in a close-to-zero-dB M det(R) term. On the other hand, the drop in η mux is caused mainly by the M M i=1 η i term, which gets closer to the η mux as the M det(R) term gets closer to 0 dB, especially in the UB.
The highest η mux is seen when the antenna is operated in free space, where η mux is between −4.07 dB and −2.17 dB in the LB, and from −1.46 dB to −0.42 dB in the UB. However, when the device is held by the user in different interaction scenarios, power is absorbed by the body tissues, hence deteriorating η mux . In the one-hand mode, η mux is reduced by around 3.05 dB and 3.81 dB in LB and UB, respectively compared to in free space. In the two-hand mode, the η mux values are 2.77 dB and 2.68 dB below the free space levels in the LB and the UB, respectively. The highest η mux loss is observed when the device is used in call mode, where the device is surrounded by more body tissues compared to the two previous interaction scenarios. The reduction in the η mux in this case is around 5.93 dB in the LB and around 5.69 in the UB relative to free space.

VI. CAPACITY
Assuming that the channel state information (CSI) is available at the receiver only, and the transmitter power is allocated uniformly cross transmit antennas, MIMO capacity can be calculated from the following relation [36]: where E is averaging operator, det is the determinant, I N r is the identity matrix of size N r where N r is the number of receive antenna elements, SNR is the signal to noise ratio, N t is the number of transmit antenna elements, H ∈ C N r ×N t is the wireless channel matrix, and () † is the transpose of the complex conjugate. The wireless channel matrix H in this work is constructed based on the model used in [37], [38]. The transmit side in this model assumes ideal antenna elements with 100 % efficiency and null ECC. In addition, the number of transmit antenna elements is fixed at 14 for all ergodic capacity calculations in this work. The wireless channel matrix H in this model is constructed as follows: where H IID ∈ C N r ×N t is the IID channel matrix which entries are complex and normally-distributed, with zero mean and unity variance specifically CN (0, 1). This matrix models the uniform propagation environment between the transmitter and the receiver. R ∈ C N r ×N r is the receiver correlation matrix which involves the effects of the AUT, as follows: where is the diagonal efficiency matrix written as: where η i is the efficiency of the i th antenna port, andR is the complex correlation coefficient matrix in (5) of the multiplexing efficiency. Furthermore, all capacities are calculated by averaging 300,000 channel realizations at every frequency point to ensure their convergence to accurate values, considering the large number of antenna elements. In addition to that, all capacity results are obtained at the SNR value of 20 dB. Fig. 10 shows the ergodic capacity of the 14 × 14 MIMO antenna system in free space and in the three interaction scenarios with human body. In Fig. 10(a), the ergodic capacity performance is higher in the UB due to the higher efficiency of antenna elements and the lower ECC 2090 VOLUME 10,2022 in this band. Moreover, the capacity changes in frequency in both bands behave similarly with the efficiency levels of the elements presented in Fig. 6. This is another indication that the wireless channel model is mainly determined by the efficiency matrix , with significantly lower contribution from the complex ECC matrixR.
In free space, the LB capacity is between 60.6 and 68.3 bit/s/Hz, while the capacity in the UB is from 70.5 to 75.0 bit/s/Hz. The highest capacity value in the UB is 2 bit/s/Hz lower than the 14 × 14 IID capacity. Meanwhile, the capacity reduction of the antenna in the presence of the human body again depends on the interaction scenario between the antenna and the body. Again, the least capacity is obtained when the mobile is used in call mode. In one-hand grip and compared to the free space performance, the LB capacity is reduced by around 12 bit/s/Hz and UB capacity is lowered by around 15 bit/s/Hz. Next and compared to the free space, the two-hand grip capacity is decreased by around 11 bit/s/Hz in both bands, and finally in the call mode is decreased the most by about 22 bit/s/Hz in both bands. Fig. 10(b) illustrates the capacity loss resulting from the interaction with the human body in percentage, relative to the free space capacity. The one-hand grip causes capacity loss between 17 and 21 % in the LB, and around 21 % in the UB. Secondly, the two-hand grip reduces the capacity from 14 to 21 % relative to free space in the LB, and around 14% in the UB. Finally the highest loss observed is in the call mode, which ranges between 31 and 38 % in the LB and around 31 % in the UB. Table 2 compares the performance of the proposed antenna array with various 5G MIMO handset antennas operating in the sub-6 GHz bands, ranging from 6 to 18 elements. It is evident from the table that the MIMO antenna performance degrades resulting from interactions with hands in various modes, as indicated by the ECC and efficiency. Furthermore, this table also highlights the number of antenna elements in each design which efficiency fell below 20 % due to severe hand blockage. Due to this, an antenna selection method is proposed to alleviate the issue of human body blockage, which will be explained in the following subsections. The following section suggests using an antenna selection to select antenna elements in good condition while excluding those that are severely obstructed by hand and head.

VII. ANTENNA SELECTION
The large number of antenna elements on the mobile terminal requires more system resources (RF circuitry, signal processing and channel estimation). Therefore, all antenna elements must effectively contribute to the system performance while consuming these resources. This section aims to understand how the different interaction scenarios between the antenna and human body can lower the contribution of selected elements in terms of capacity.
Finally, an antenna selection algorithm is used to optimize the overall contribution of these antenna elements in terms of capacity. In this section, the number of transmit antennas N t is fixed at 14, and on the receiver side L antennas are selected from the N r available antennas on the mobile terminal. Furthermore, the wireless channel matrix with the antenna selectionH ∈ C L×N t is constructed by selecting L rows from the available rows N r in the channel matrix H of the full system. Hence, the set S which contains all possible differentH matrices with L rows (i.e. L receive antennas) has a cardinality of M where M = N r L .
The optimal matrixH opt must be selected among all possible M matrices in S to maximize the system capacity as follows [44]: Selection of theH opt can be conducted by an exhaustive search method, where the capacity from all possible M matrices in the set S is computed before the matrix with the highest capacity is selected. However, computing the capacities from all possible matrices in the set S is computationally complex and resource-consuming, especially with increasing number of antennas on the device. Therefore, several algorithms have been proposed to select the optimalH opt with a balance of VOLUME 10, 2022 computational complexity and losses in performance [45]. Practical implementation of antenna selection algorithms can be hard when a set of switches is used to activate selected antennas only, or a soft selection can be implemented when all antennas are kept active and selection algorithm is implemented in the RF domain upon the received signals from all antennas [6]. In this work, an algorithm to select the L receive antennas on the mobile terminal (and consequently the optimal matrixH opt ) is proposed based on the received power from each receive antenna. The i th row in the full system channel H represents channel links from all transmitter antennas to the i th receiver antenna. Thus, the Frobenius norm I is taken for each row as a measure of the received power through each receive antenna. Then, the L rows corresponding to the highest I values are selected to build the channel matrixH opt . The Frobenius norm I i of the i th row for 1 ≤ i ≤ N r is calculated as follows [46]: (11) where h ij is the (i, j) th entry of the full system channel matrix H.
The capacity with the antenna selection algorithm is presented in Fig. 11. Besides the 14 × 14 system capacity, the figure also shows the capacity of antenna selection algorithm with three different L values of receive antennas (12, 10 and 8 elements). Results obtained in free space and under the influence of the human body in the three interaction scenarios are also presented.
When the two least-contributing elements are removed and the remaining 12 elements considered, the free space capacity is reduced by 5.4 bit/s/Hz in the LB and 6.2 bit/s/Hz in the UB compared to the full system capacity. Despite that, a smaller capacity reduction is observed with the 12 antenna elements when the device is used in the vicinity of human body. The capacity decreased from 2.3 to 2.7 bit/s/Hz below the full system capacity in both frequency bands and in the three cases of the interaction with the human body. Next, 10 antenna elements are selected to be evaluated. The capacity reduction in free space is from 12.4 to 14.5 bit/s/Hz in both frequency bands. On the other hand, there is less reduction observed when the device is used in the vicinity of the human body. The capacity declined by 6.8 to 8.7 bit/s/Hz in the three interaction cases over both frequency bands. Finally, when only 8 antenna elements are selected, the free space capacity is lowered by 20.8 to 24.2 bit/s/Hz in both bands. Moreover, the capacity in the vicinity of the human body declines by up to 12.0 bit/s/Hz in the call scenario and up to 16.8 bit/s/Hz when the device is held by two hands. It is seen that the excluded elements in all cases contributed minimally to the capacity when the device is used in the vicinity of human body compared to in the free space. This reduced contribution is due to the low performance of these excluded elements due to body blockage. However, note that the capacity drop in Fig. 11 is discussed as the difference in bit/s/Hz unit between the full system capacity and the capacity of different L-element system. This is without considering the peak values of the full MIMO antennas' capacity. Fig. 12 shows the contribution of the L-element MIMO antenna in terms of capacity compared to the capacity of the full system. When L = 12, the free space capacity is up to 91.7 % relative to the 14-element free space capacity. This is then increased from 94.0 % to 96.3 % when the mobile terminal is evaluated in the vicinity of the human body for each case. Next, with L = 10, the system achieves up to 80 % of the 14-element antenna's free space capacity, whereas this increased to above 85 % in the vicinity of human body in the three different interaction scenarios. Lastly, the 8-element MIMO antenna's capacity achieved up to 67 % of the full system capacity in free space. This capacity then increased to above 72 % in the vicinity of the body.
From the results in Fig. 11 and Fig. 12, a selection algorithm can be applied to exclude less-contributing elements, especially when the device is used in the vicinity of the human body. This is to avoid the consumption of the system resources by these elements while they contribute less significantly in the performance. Therefore, exclusion of these elements enables the optimization of system resources with acceptable deterioration in the system capacity. On the other hand, the mobile terminal can be designed with extra antenna elements on the device, and the algorithm can periodically search for the best elements to be used especially when it is used in the vicinity of the human body. It is understood from the results in the previous sections that the least-performing elements (which are excluded from VOLUME 10, 2022 the channel matrix) change depending on the interaction scenarios with human body. Fig. 13 shows the probability of each element to be selected when 10 elements are considered. In free space, it is observed that all elements are equally likely to be selected due to the similar efficiency of antenna elements in free space, as shown in Fig. 6. However, when the device is used in the vicinity of the human body, the likelihood of an element being selected depends mainly on its interaction with the body, besides the quality of the propagation link of the particular channel realization.

VIII. CONCLUSION
This work presented a study on a 14-element MIMO antenna designed to operate for 5G applications. The antenna performance is first evaluated in free space before being assessed when used in three different near-human body cases. Despite maintaining the ECC values of less than 0.3 in all evaluation cases, the efficiency of the different antenna elements varies considerably, subjected to the type of interaction scenario with the body. The performance of the multiplexing efficiency and ergodic capacity showed that the antenna efficiency limitations are the main factor in the performance degradation. In addition to that, the antenna selection results indicated that the contribution of elements in the channel can be predicted from the antenna design stage. Further work is needed to further quantify the tradeoff between the contribution of antenna elements under the influence of the human body and the resources allocated to these elements.