**3. Channel measurements and models**

The measurements were carried out using an Agilent 8270ES vector network analyzer (VNA). The antennas used were SkyCross SMT-3TO10M-A. These antennas are linearly polarized and azimuthally omnidirectional. The cables used were 5 m long SUCOFLEX RF with 7.96 dB loss. The data acquisition system included a computer with LabVIEWTM software. The VNA was operated in a transfer function measurement mode, where Port 1 and Port 2 are the transmitting and receiving ports respectively, as seen in Fig. 4.

**Figure 4.** Measurement setup

This setup corresponds to a measurement of the S21 parameter where the device under test (DUT) is the radio channel. The range of the frequency spectrum covered was from 3.1 GHz to 10.6 GHz. For each experiment setup 100 frequency responses were measured. The measurement parameters are summarized in Table 2.


**Table 2.** Measurement parameters

88 Ultra Wideband – Current Status and Future Trends

Regular Room Subject Standing

**Table 1.** Measurements and scenarios.

**Figure 4.** Measurement setup

bed

**3. Channel measurements and models** 

Surgery Room

**Environment Scenarios** 

2 people walking around the

The measurements were carried out using an Agilent 8270ES vector network analyzer (VNA). The antennas used were SkyCross SMT-3TO10M-A. These antennas are linearly polarized and azimuthally omnidirectional. The cables used were 5 m long SUCOFLEX RF with 7.96 dB loss. The data acquisition system included a computer with LabVIEWTM software. The VNA was operated in a transfer function measurement mode, where Port 1

and Port 2 are the transmitting and receiving ports respectively, as seen in Fig. 4.

Subject Lying Links A1 and A2

Subject Lying

Links A1 and A2

2 people walking around the bed, one of them is using a mobile phone

Links A1 and A2

Links A1 and A2

Subject Lying

The measured transfer function frequency values were converted to the time domain (channel impulse response) using an inverse Fast Fourier Transform. A Hamming window was used to reduce sidelobes.

## **3.1. Channel impulse response**

Fig. 5 shows the average of the channel impulse response, corresponding to link A1, when the subject is standing in the hospital room shown in Fig. 2.

The effect of the human body and the environment can be clearly differentiated. These results are significantly different than the ones obtained in an empty hospital room (Hentilä et al, 2005). In Fig. 5 the first region of the IR shows a fast decay of the energy during the first 5-6 ns due to the effects of the human body. The decay of the second region in the response is slower and contains the diffuse multipath components and a few subclusters caused by the reflections coming from the room. In this particular case the first of such subclusters, arriving at around 8 ns, is due to a measuring equipment (VNA) which was located 1.3 m in front of the subject when is standing.

For each particular hospital scenario listed in Table 1 the measurements obtained share the general characteristics shown in Fig. 5.

Fig. 6 corresponds to the case when the subject is lying down on bed in a hospital room. Fig. 7 corresponds to the case when the subject is lying down on a bed in a surgery room and two other people are randomly walking around the bed. The Least Squares (LS) fitted lines shown in these figures are used to model the variability of the amplitudes as described in Section 3.2.

The UWB Channel in Medical Wireless Body Area Networks (WBANs) 91

��� ) (1)

**Figure 7.** Average channel impulse response of the radio link A1. The subject is lying down on a bed in

Once a set of measurements have been obtained they are used to estimate the parameters of a common mathematical representation of communications channels, a tapped delay line.

ℎ(�) = ∑ ���(����)����(���

In equation (1) � is the number of paths. For the �th path, �� is the signal amplitude, �� is the arrival time, and �� is the phase. For the case shown in Fig. 5 (when the subject is standing in a regular hospital room) it is more appropriate to consider two regions for the modeling of the signal clusters, each with it is own set of distributions for the characterization of the

As illustrated in Fig. 8, the values of �� in equation (1) can be approximated by two exponential decaying functions, one for each region. Using a least squares (LS) method these functions (when expressed in dB) are best fitted with a Rician factor �� and an exponential

���

a surgery room and two other people are randomly walking around the bed.

**3.2. Channel models** 

amplitudes decays and intearrival times.

*3.2.1. Exponentially decaying factor* 

decaying factor Γ (equation (2)).

**Figure 5.** Average channel impulse response of the radio link A1. The subject is standing in a regular hospital room.

**Figure 6.** Average channel impulse response of the radio link A1. The subject is lying down on a bed in a regular hospital room.

**Figure 7.** Average channel impulse response of the radio link A1. The subject is lying down on a bed in a surgery room and two other people are randomly walking around the bed.

#### **3.2. Channel models**

90 Ultra Wideband – Current Status and Future Trends

hospital room.

a regular hospital room.

**Figure 5.** Average channel impulse response of the radio link A1. The subject is standing in a regular

**Figure 6.** Average channel impulse response of the radio link A1. The subject is lying down on a bed in

Once a set of measurements have been obtained they are used to estimate the parameters of a common mathematical representation of communications channels, a tapped delay line.

$$h(\tau) = \sum\_{l=0}^{L-1} a\_l \delta(\tau - \tau\_l) \exp(j\phi\_l) \tag{1}$$

In equation (1) � is the number of paths. For the �th path, �� is the signal amplitude, �� is the arrival time, and �� is the phase. For the case shown in Fig. 5 (when the subject is standing in a regular hospital room) it is more appropriate to consider two regions for the modeling of the signal clusters, each with it is own set of distributions for the characterization of the amplitudes decays and intearrival times.

#### *3.2.1. Exponentially decaying factor*

As illustrated in Fig. 8, the values of �� in equation (1) can be approximated by two exponential decaying functions, one for each region. Using a least squares (LS) method these functions (when expressed in dB) are best fitted with a Rician factor �� and an exponential decaying factor Γ (equation (2)).

$$\begin{aligned} 10\log\_{10}|a\_l| &= \begin{cases} 0, & l=0\\ \mathcal{Y}\_{01} + 10\log\_{10}\left(\exp\left(\frac{-t\_l}{\Gamma\_1}\right)\right), & 1 \le l \le l\_1\\ \sum\_{m=1}^{M} \left(\mathcal{Y}\_{02m} + 10\log\_{10}\left(\exp\left(\frac{-t\_l}{\Gamma\_{2m}}\right)\right)\right), & l\_2 \le l \le L-1 \end{cases} \end{aligned} \tag{2}$$

$$p(t\_l|t\_{l-1}) = \begin{cases} \lambda\_1 \exp\{-\lambda\_1(t\_l - t\_{l-1})\}, & 1 \le l \le l\_1\\ \lambda\_2 \exp\{-\lambda\_2(t\_l - t\_{l-1})\}, & l\_2 \le l \le L - 1 \end{cases} \tag{3}$$

$$p(L) = \frac{\mu\_L \exp(\mu\_L)}{L!} \tag{4}$$

**Figure 10.** Magnitude of the channel impulse response for each position of the walking cycle.

These experimental results indicate that the arm movements have a significant impact on the radio link A1 (Tx antenna on the left wrist and Rx antenna in the center of the front torso). For instance, when the hand moves to position three the strongest path arrives earlier than in the other positions due to the shorter distance between the antennas. There are also more significant paths due to the interaction of the electromagnetic waves with of the arm and the shoulder. The shadowing of the signal due to blocking by the body is evident in position six, where the left hand moves to the lowermost location.

Fig. 11 provides an alternative view of the channel impulse response where the delay of the most prominent peak is clearly shown (Taparugssanagorn et al. 2011).

To evaluate the delay dispersion within the channel the root mean square (RMS) delay spread ���� is estimated. The ���� is defined as,

$$\tau\_{RMS} = \sqrt{\frac{\sum\_{l=0}^{L-1} (\tau\_l - \tau\_m)^2 |h(\tau\_l)|^2}{\sum\_{l=0}^{L-1} |h(\tau\_l)|^2}} \tag{5}$$

The UWB Channel in Medical Wireless Body Area Networks (WBANs) 95

**Figure 11.** Channel impulse response (link A1) for each position of the walking cycle.

Access Protocols (MAC) to be used in medical applications (Viittala et al., 2009).

**5. UWB radar in medical applications** 

infer the nature of the implant behaviour.

BroadSpecTM UWB antennas were used for this experiment.

The results presented in this and in the previous section highlight the importance of properly understand and model the UWB channel when designing physical and Medium

The potential use of UWB technology goes beyond transmitting information, collected by sensors, to a control station. The nature of the UWB signal is such that it can be used as in common radar applications, e.g. to detect and estimate dynamic parameters of an object. Fig. 12 shows the channel impulse responses for the case of subjects with and without an aortic valve implant (Taparugssanagorn et al. 2009). The Rx antenna was located at the middle of the front torso and the Tx antenna close to the heart, 10 cm away from the Rx antenna. P200

It apparent that the responses are different, i.e. the one corresponding to the subject with an aortic implant has lower peaks. A possible explanation for the difference in the responses is the scattering caused by the metallic (titanium alloy) valve. Subsequent simulation studies carried out using a 3D immersive visualization environment has confirmed this type of results (Yang et al., 2011). Further investigations could lead to the use of the response to

The use of UWB signals to directly monitor vital signs is currently a very active research area. Thus for example, the estimation of the breathing rate and the heart beat frequency has been studied in (Lazaro et al., 2010). Using a mathematical model of the human body as

where �� is the mean excess delay defined as,

$$\tau\_m = \frac{\sum\_{l=0}^{L-1} \tau\_l |h(\mathbf{r}\_l)|^2}{\sum\_{l=0}^{L-1} |h(\mathbf{r}\_l)|^2} \tag{6}$$

�(�) is the channel impulse response, L is the number of paths and � is the delay. For the case discussed here the estimates for mean and the standard deviation of ���� are 0.1371 ns and 0.0670 ns respectively. Also the probability distribution function that best fits the variations of the amplitude is the Weibull distribution.

**Figure 11.** Channel impulse response (link A1) for each position of the walking cycle.

The results presented in this and in the previous section highlight the importance of properly understand and model the UWB channel when designing physical and Medium Access Protocols (MAC) to be used in medical applications (Viittala et al., 2009).
