3.1. Network Model
We consider network deployment in the hospital with two main e-health applications, telemedicine and hospital information systems, as in [
5,
10]. The telemedicine application involves providing real-time healthcare service delivery to distant users, such that the sensor nodes send the vital signal to the coordinator, and the coordinator transmits those signals to the gateway or controller [
5]. The hospital information system collects the data of patients at the hospital. Because the telemedicine application has a higher priority than the hospital information application, users of telemedicine application are considered as PUs and user of hospital information application is considered as SUs. We assume that PUs will turn on within a specific duration of time according to the needs of patients. In addition, on taking into consideration the traffic priority in the WBAN, we also consider the WBAN with the highest traffic priority in
Table 1 as the PU. The notations of the network model and scheduling problem are shown in
Table 2.
In order to evaluate the SPLS algorithm, we assume that the three-tier CRBAN is set up as in
Figure 1. The particular network scenario of our proposed work in
Figure 2 is similar to the hospital scenario in [
5,
10], which is divided into nine similar rooms, where each room has an area of 9 m
2. The network consists of a controller and multiple CRBANs with the LS and NLS medical devices. Each CRBAN or SU comprises several sensors on the human body, which is allowed to move freely through the area. In each room, the locations of the NLS and LS medical devices are fixed, and the CRBANs are uniformly distributed in the area. The movement of CR clients is modeled as a random mobility model as in [
10].
Taking into consideration the priority of various healthcare services, we define the priority level of CRBANs as follows. The CRBANs can be PUs or SUs depending on the healthcare services as listed in
Table 1. In the case of a licensed channel, the CRBANs are SUs with transmissions scheduled according to the CRBAN services. In the case of the unlicensed channel, the nodes for the CRBAN service with the highest priority in
Table 1 are regarded as PUs, and the nodes for other CRBAN services are regarded as SUs.
The CRBANs use three kinds of channels called CtrlCh channel, DataCh channel, and EDataCh channel. The CtrlCh is used for transmission of control signals between the coordinators and sensor nodes or between the coordinators and the controller. For example, the CtrlCh is used for the broadcast packet and the beacon packet of the coordinator. The EDataCh is only used for transmission of services with the highest priority shown in
Table 1, while the DataCh is used for transmission of normal services. In this system model, CRBANs use the control channel to exchange messages with each other, and the beacon signal of the CRBAN is also transmitted on the control channel when the coordinator senses the idle control channel. The sensor nodes wait for the beacon packet on the control channel and then switch to the data channel for data transmission according to the information in the beacon signal. The CRBAN accesses the control channel by using carrier-sense multiple-access/collision-avoidance (CSMA/CA) protocol. The overlay spectrum is used as an interference mitigation model where the CRBANs use the spectrum that is not occupied by the PUs. The medical devices work on either licensed or unlicensed channels.
The spectrum is divided as follows: the licensed band is specified for wireless medical telemetry service in the spectrum bands 608–614 MHz, 1395–1400 MHz, and 1427–1432 MHz and medical implant communications service in the spectrum band 402–405 MHz; the unlicensed band is the ISM band at 2.4 GHz [
5,
10]. In [
3], as per the IEEE standard 802.15.6, the narrowband band 2400–2483.5 MHz is divided into 79 channels of 1-MHz bandwidth. However, we only consider 20 unlicensed channels at 2.4 GHz for simplicity. In the unlicensed band, the frequency 2483 MHz is selected for CtrlCh because it is not overlapped with the IEEE 802.11 channels, and the frequency 2475 MHz is selected for EDataCh, while DataCh is selected as any idle channel in the ISM band (18 channels). The licensed band is used for the medical devices or PUs, CRBANs sense the vacant channel or the spectrum hole.
We consider two types of transmission: the transmission between the sensor nodes and the coordinator as an intra-CRBAN transmission and the transmission from the coordinator to the controller or the transmission from the coordinator of one CRBAN to the coordinator of other CRBANs as an inter-CRBAN transmission.
In the 2.4–2.5-GHz band, the channel model for intra-CRBAN follows a power law model as per the IEEE standard 802.15.6 as in [
18], and the path loss is calculated by:
where
d is the distance between the transmitter and receiver in mm,
a (6.60) and
b (36.1) are parameters of the model, and
NPL is a normally distributed variable with standard deviation σ
N of 3.80.
The inter-CRBAN channel model is considered as a distance-dependent path loss model, the path loss exponent is less than two, and the fading follows a gamma distribution. The mean and variance values follow a power law with respect to the distance between CRBANs, in which the rate of increase of path loss depends on the increase of the distance between two CRBANs [
18,
19]. In [
18,
19], the measurement of inter-CRBAN channel model is modeled as the classical distance-dependent path loss model as follows:
where
Pgain is either the mean or variance of the path gain,
rdB = 20log
10r, where
r is the distance between two CRBANs, and
a and
b are the fitting parameters at 2.45 GHz, respectively. The mean of
a and
b is −0.05 and −0.19, respectively, and the variance of
a and
b is −0.19 and −52.8, respectively.
The path loss between the CRBAN and controller is considered as an indoor path loss model. The transmit power of the medical devices is similar to that in [
5,
10], and thus the total path loss of a CRBAN in the hospital environment is calculated as follows:
where
d is the distance from the medical devices to the CRBAN, and
n is the number of floors (or walls) the radio signal has to traverse.
The upper bound on the transmit power for NLS and LS devices are defined in [
5,
8] as follows:
where
DNLS(
n) and
DLS(
m) are the distances from the CRBAN to the NLS and LS devices, respectively.
ENLS(
n) and
ELS(
m) are the electromagnetic interference immunity levels for the NLS and LS medical devices
n and
m, respectively, in
Figure 2.
Because the NLS and LS devices operate in the same vicinity of CRBANs, the interference probability was defined if the CRBAN causes interference with the medical devices by violating the transmit power constraints
PNLS and
PLS. As in [
5,
10], the transmit power in the data channel allows a successful transmission from the CRBANs to the controller, which is defined as:
where
PNLS and
PLS are derived from (4) and (5), respectively.
In [
10], in the area with a high electromagnetic interference (EMI) level (mainly due to a large number of life-supporting medical devices), the CRBAN cannot reach the controller with the minimum required signal. In our overlay algorithm, the CRBANs change the operating channel such that they can transmit with a high power while causing no interference with medical devices.
3.2. Channel Sensing
The coordinator performs channel sensing on entire licensed channels and unlicensed channels. The coordinator records the signal to interference plus noise ratio (SINR) of each channel in order to determine whether there is any transmission of PUs on a channel. The SINR observed at the coordinator of
CBi in channel
Ck is defined as:
where
Pr is the received signal,
Ik is the interference power in channel
Ck, and
N0 is the additive white Gaussian noise.
In a practical scenario, the value of SINR can be estimated at the PHY layer of the receiver as in [
20] using the RSSI (received signal strength indicator) as follows:
where
η0 is the thermal noise, the constant
C is the measurement offset that is empirically measured in [
20] using Chipcon CC2420 on the Telos motes (
C = 45 dB), and the value of 30 is the conversion of dBm to dB.
As in [
21,
22,
23], channel
Ck is idle if there is no PU activity on the licensed channel or no transmission on the unlicensed channel. In such a case, the result of channel sensing at the coordinator of
CBi on channel
Ck is denoted as:
We assume that the PU user-activity models are similar to those in [
21,
22,
23]. The PU’s user-activity model has two states, which are
Idle and
Busy, as shown in
Figure 3. The values
p and
q are the probability that an idle channel becomes busy and the probability that a busy channel becomes idle, respectively. The durations of the busy and idle times of the PU are defined as
Tbusy(
k) and
Tidle(
k), respectively. The arrival of the PU is independent of CRBANs’ activities, and the transition follows a Poisson process in which the lengths of both periods are exponentially distributed with rate λ and mean value
E = 1/λ. If
q >
qthres or (1 –
p) >
pthres, then the coordinator estimates an idle channel
Ck at CRBAN
CBi is obtained as:
where Δ
Tt(
Ck) = 1 indicates that
Ck is idle during one superframe. The coordinator evaluates the possible time that the duration Δ
Tk for which the
CBi can occupy
Ck is longer than a threshold duration
Tthres, which is equal to the length of a superframe.
The list of idle channels for the data transmission that is observed by the coordinator is denoted as:
The coordinator broadcasts the list CIi(t) on CtrlCh to the network to discover the nearby CRBANs. The neighboring discovery and link scheduling steps are explained in the next section.