**3. Signal processing**

### **3.1. Energy detection for UWB communications**

For UWB communications, the detection of transmit symbols can be done with a coherent or a non-coherent method. Typically, with perfect synchronization, the coherent detection based on correlation gives a better performance than non-coherent detection. Coherent detection is known to be the optimum method with respect to the bit error rate (BER) for AWGN channels. In case of multipath propagation channels, the transmit impulse and the channel impulse response together (convolved) must be used for correlation detection, and in general more complex signal processing is required. Analog signal processing is considered here because analog to digital conversion for UWB signals is hardly available and requires high power consumption. Unfortunately, the coherent approaches are not well suited for our large bandwidth analog signal processing within the receiving frontend. Energy detection is usually preferred because, if applied in a proper way, no channel impulse response is needed in the receiver. Moreover, energy detection is also robust with respect to synchronization accuracy. In its basic form the energy detector consists of a squaring device, an integrator, a sampler and a decision device.

Pulse position modulation (PPM) and On-off keying (OOK) are modulation techniques that are typically used in combination with energy detection. The data modulation is performed by changing the position or amplitude for PPM and OOK respectively. The detection of PPM is easier to perform since comparing the signal energy at two different intervals is enough, while OOK requires threshold estimation. We look at the BER performance of PPM with respect to synchronization and multipath propagation. Different levels of synchronization in AWGN channels (perfect, ±20 ps, and ±40 ps) and a perfect synchronization for a multipath channel are considered. The errors in the synchronization accuracy are uniformly distributed. The transmitted signal is an impulse train of fifth derivative Gaussian functions (*σ* = 51 ps) with PPM modulation. The tested channels AWGN channel and a multipath channel with delay spread are assumed to be shorter than 4 ns, which is half of the impulse repetition interval. The correlation receiver uses a template impulse (fifth derivative Gaussian) for correlation, and the correlation is centered at the first strongest path of the channel. The BER for all settings is shown in Fig. 22.

**Figure 22.** BER performance of PPM modulation for correlator and energy detection with AWGN channel and multipath channel.

The performance of the correlation receiver suffers from synchronization uncertainty and multipath propagation. This is due to the fact that the impulse used in UWB systems is very narrow and the impulse correlation receiver cannot capture all of the signal energy. On the other hand, the energy detector shows good performance also for non-perfect synchronization and multipath channel. We can conclude from the results that, with a performance trade-off, the energy detection is much more robust. Other challenges for implementing energy detection are multiuser capability and interference cancellation. These problems can be solved by the comb filter receiver presented in the following part.

### **3.2. Comb filter**

18 Will-be-set-by-IN-TECH

**Figure 21.** Real time oscilloscope trace showing the functional test of the monostatic radar frontend,

For UWB communications, the detection of transmit symbols can be done with a coherent or a non-coherent method. Typically, with perfect synchronization, the coherent detection based on correlation gives a better performance than non-coherent detection. Coherent detection is known to be the optimum method with respect to the bit error rate (BER) for AWGN channels. In case of multipath propagation channels, the transmit impulse and the channel impulse response together (convolved) must be used for correlation detection, and in general more complex signal processing is required. Analog signal processing is considered here because analog to digital conversion for UWB signals is hardly available and requires high power consumption. Unfortunately, the coherent approaches are not well suited for our large bandwidth analog signal processing within the receiving frontend. Energy detection is usually preferred because, if applied in a proper way, no channel impulse response is needed in the receiver. Moreover, energy detection is also robust with respect to synchronization accuracy. In its basic form the energy detector consists of a squaring device, an integrator, a sampler and

Pulse position modulation (PPM) and On-off keying (OOK) are modulation techniques that are typically used in combination with energy detection. The data modulation is performed by changing the position or amplitude for PPM and OOK respectively. The detection of PPM is easier to perform since comparing the signal energy at two different intervals is enough, while OOK requires threshold estimation. We look at the BER performance of PPM with respect to synchronization and multipath propagation. Different levels of synchronization in AWGN channels (perfect, ±20 ps, and ±40 ps) and a perfect synchronization for a multipath channel are considered. The errors in the synchronization accuracy are uniformly distributed. The transmitted signal is an impulse train of fifth derivative Gaussian functions (*σ* = 51 ps) with PPM modulation. The tested channels AWGN channel and a multipath channel with delay spread are assumed to be shorter than 4 ns, which is half of the impulse repetition interval. The correlation receiver uses a template impulse (fifth derivative Gaussian) for correlation, and the correlation is centered at the first strongest path of the channel. The BER for all settings

displaying transmitted pulse crosstalk and received pulse echo.

**3.1. Energy detection for UWB communications**

**3. Signal processing**

a decision device.

is shown in Fig. 22.

The received signal power for medical applications are expected to be very small due to high attenuation in human tissue. We propose a receiver based on a comb filter to improve Signal-to-Noise ratio (SNR) before further processing. The comb filter is a feedback loop with an analogue delay and a constant loop gain of one for all frequencies. It is used to perform a coherent combination of the incoming UWB impulses. The feedback loop sums up the number of impulses used for the transmission of a data symbol/measurement and is reset after this. The coherent combination results in SNR improvement, interference suppression which come from different transmitters in a multiuser environment or narrowband interfering signals. Several UWB impulses are transmitted for one data symbol/measurement. One important feature of the concept is that the individual UWB impulses are weighted by +1 or -1 according to a spreading sequence. The UWB transmit signal *s*(*t*) can be written as

$$s(t) = \sum\_{k=-\infty}^{\infty} \sum\_{n=0}^{N-1} c\_n p(t - nT\_c - kT\_s),\tag{6}$$

where *p*(*t*) is a UWB impulse and *cn* is the spreading sequence. *Tc* is the period between two UWB impulses or 'chip period' and *Ts* = *NTc* is symbol period for communications or measurement period for radar/localization application. For communication, assuming a binary transmission, the impulse train of each data symbol can be modulated in different

#### 20 Will-be-set-by-IN-TECH 458 Ultra-Wideband Radio Technologies for Communications, Localization and Sensor Applications UWB in Medicine – High Performance UWB Systems for Biomedical Diagnostics and Short Range Communications <sup>21</sup>

ways. We consider three modulations using the direct sequence spread spectrum technique based on OOK, PPM, Code shift keying (CSK) ,i.e. DS-OOK, DS-PPM and DS-CSK. For DS-OOK, transmitting a train of impulses represents the data '1', while no impulse signals means '0'. For DS-PPM, the two basic waveforms for a binary transmission are different by time shift. For DS-CSK, the two waveforms result from two different spreading sequences. A decision threshold is not required for DS-PPM and DS-CSK which is a big advantage compared to DS-OOK. The basic waveforms for different modulation techniques are illustrated in Fig. 23.


**Figure 24.** Block diagram of the basic comb filter based receiver.

and hence it could violate the spectral mask requirement. The longer and more random the spreading sequence, the better. Of course, a direct restriction is given by the desired data rate.

459

UWB in Medicine – High Performance UWB Systems for Biomedical Diagnostics and Short Range Communications

In general we will have many transmitters or many sensors and different corresponding spreading sequences. The sequences are selected such that the mutual cross-correlation values are as small as possible. This guarantees that many transmitters can be used at the same time in the same area, e.g. several implants for digital transmission. The signals from all sensors are assumed to be transmitted in parallel with synchronization in a certain time window. The remodulation and the comb filter at the receiver is the crucial part in suppressing the multiuser interference (MUI). The weighting factors of the channel impulse response train after the multiplication are a scrambled version between transmit and receive sequence, if they are not the same. The accumulation at the comb filter would eliminate the MUI and the result of the weight is the cross-correlation values. The illustration of multiplication with the spreading sequence and comb filter accumulation of a signal with two different sequences is

**Figure 25.** Comb filter principle in the receiver.

*3.2.2. Multiuser interference*

shown in Fig. 26.

**Figure 23.** Comparison of UWB transmit signals for different modulation techniques.

Transmitted through the channel, the impulses are affected by the channel impulse response which is expected to be a multipath environment with or without a line-of-sight path. The chip period is chosen to be larger than the multipath spread, with the result that no interchip interference (i.e. no overlapping of channel impulse responses) occurs. We assume the channel to be time invariant within the symbol period. This means that the signal or basic waveform which represents one data symbol/measurement in the received signal consists of a corresponding number of channel impulse responses.

### *3.2.1. Basic comb filter receiver*

The receiver based on the comb filter with remodulation is shown in Fig. 24. The receiver consists of an antenna, a LNA, a multiplier, the comb filter, the energy detector for communications and a correlator for radar/localization applications. This concept allows energy detection in a multipath and multi-sensor environment and amplification of the impulse response in the comb filter delay loop.

In the receiver, the received signal is multiplied with the spreading waveform (i.e. the sequence with rectangular "chips"). If the spreading sequence matches, the result is a periodically repeated channel impulse response with period *Tc*. After multiplication, the received impulses are delayed and summed up by the comb filter. At the output of the comb filter, only the last *Tc* period of every *Ts* is used for further processing which is expected to be an amplified and improved SNR channel impulse response. Interference from different UWB transmitters and other systems are eliminated at the comb filter. The process of remodulation and coherent combination at the comb filter is illustrated in Fig. 25.

Spreading is used not only for the multiuser or multi-sensor purpose, but also for shaping the power spectral density of the transmitted signal. The impulse train may have high spikes in the power density spectrum, because it has a periodic behavior over a longer time interval,

**Figure 24.** Block diagram of the basic comb filter based receiver.

**Figure 25.** Comb filter principle in the receiver.

and hence it could violate the spectral mask requirement. The longer and more random the spreading sequence, the better. Of course, a direct restriction is given by the desired data rate.

### *3.2.2. Multiuser interference*

20 Will-be-set-by-IN-TECH

ways. We consider three modulations using the direct sequence spread spectrum technique based on OOK, PPM, Code shift keying (CSK) ,i.e. DS-OOK, DS-PPM and DS-CSK. For DS-OOK, transmitting a train of impulses represents the data '1', while no impulse signals means '0'. For DS-PPM, the two basic waveforms for a binary transmission are different by time shift. For DS-CSK, the two waveforms result from two different spreading sequences. A decision threshold is not required for DS-PPM and DS-CSK which is a big advantage compared to DS-OOK. The basic waveforms for different modulation techniques

**Figure 23.** Comparison of UWB transmit signals for different modulation techniques.

a corresponding number of channel impulse responses.

impulse response in the comb filter delay loop.

and coherent combination at the comb filter is illustrated in Fig. 25.

Transmitted through the channel, the impulses are affected by the channel impulse response which is expected to be a multipath environment with or without a line-of-sight path. The chip period is chosen to be larger than the multipath spread, with the result that no interchip interference (i.e. no overlapping of channel impulse responses) occurs. We assume the channel to be time invariant within the symbol period. This means that the signal or basic waveform which represents one data symbol/measurement in the received signal consists of

The receiver based on the comb filter with remodulation is shown in Fig. 24. The receiver consists of an antenna, a LNA, a multiplier, the comb filter, the energy detector for communications and a correlator for radar/localization applications. This concept allows energy detection in a multipath and multi-sensor environment and amplification of the

In the receiver, the received signal is multiplied with the spreading waveform (i.e. the sequence with rectangular "chips"). If the spreading sequence matches, the result is a periodically repeated channel impulse response with period *Tc*. After multiplication, the received impulses are delayed and summed up by the comb filter. At the output of the comb filter, only the last *Tc* period of every *Ts* is used for further processing which is expected to be an amplified and improved SNR channel impulse response. Interference from different UWB transmitters and other systems are eliminated at the comb filter. The process of remodulation

Spreading is used not only for the multiuser or multi-sensor purpose, but also for shaping the power spectral density of the transmitted signal. The impulse train may have high spikes in the power density spectrum, because it has a periodic behavior over a longer time interval,

are illustrated in Fig. 23.

*3.2.1. Basic comb filter receiver*

In general we will have many transmitters or many sensors and different corresponding spreading sequences. The sequences are selected such that the mutual cross-correlation values are as small as possible. This guarantees that many transmitters can be used at the same time in the same area, e.g. several implants for digital transmission. The signals from all sensors are assumed to be transmitted in parallel with synchronization in a certain time window. The remodulation and the comb filter at the receiver is the crucial part in suppressing the multiuser interference (MUI). The weighting factors of the channel impulse response train after the multiplication are a scrambled version between transmit and receive sequence, if they are not the same. The accumulation at the comb filter would eliminate the MUI and the result of the weight is the cross-correlation values. The illustration of multiplication with the spreading sequence and comb filter accumulation of a signal with two different sequences is shown in Fig. 26.

**Figure 26.** Illustration of multiuser interference suppression at the comb filter.

We assume synchronization in a certain time window between different users. With this assumption, we mainly consider m-sequences because they have low crosscorrelation and are simple to generate. Alternatively, binary zero-correlation-zone (ZCZ) sequence can also be considered. Unlike m-sequences, these sequences have a crosscorrelation of zero within a small window but the drawback is a smaller number of sequence in one set [40]. Binary ZCZ sequence sets were investigated in [37] for communication because of their potential to completely eliminate multi-sensor and interchip interference (ICI). A near-far problem was taken into account and we can see an advantage of this sequence set.

(a) (b)

**Figure 27.** (a) Input and output of the comb filter with remodulation. (b) BER performance of DS-PPM

For physical systems this calculation is only valid if no instability is introduced. Therefore the values of the gain *Gc* of the comb filter can only be in the range of 0-1, because otherwise an oscillation occurs. *Gc* is designed to be one but considering real components, a loss in the loop is possible. The comb filter processing gain *Gp* is a function of the gain *Gc* of the comb filter

⎛

�

<sup>∑</sup>*N*−<sup>1</sup> *<sup>n</sup>*=<sup>0</sup> *<sup>G</sup><sup>n</sup> c* �2 ⎞

*Ts* · sinc(*Ts*(*f* − *k*/*Tc*)) exp(−*jπ f Ts*) (8)

⎟⎠ (7)

UWB in Medicine – High Performance UWB Systems for Biomedical Diagnostics and Short Range Communications

461

<sup>∑</sup>*N*−<sup>1</sup> *<sup>n</sup>*=<sup>0</sup> *<sup>G</sup>*2*<sup>n</sup> c*

⎜⎝

The maximum processing gain *Gp* of the comb filter is 10 log(*N*) dB, when *Gc* = 1. The processing gain is reduced if *Gc* is less than one. The relations between *Gp*, *Gc* and *N* are shown in Fig. 28. We can see from Fig. 28(a) that the processing gain *Gp* saturated in the case where *Gc* is less than one because the impulse energy vanishes after some iterations. With a higher number of impulses per symbol, the comb filter loop gain has to be controlled more

Since UWB covers a very large bandwidth, strong interference within the band is possible and can cause problems at the receiver. Comb filter in combination with multiplication with the spreading waveform can suppress narrowband interference very well. Only a periodic signal that has a period which equals one or multiples of the comb filter delay can go through the

We can see that the transfer function of the comb filter consists of several peaks. The peaks could be seen as tunnels that allow only signals with specific frequencies to pass. After the

and the number of iterations *N*. It can be calculated as

precisely as shown in Fig. 28(b).

*3.2.4. Narrowband interference*

*Gp* = 10 · log10

comb filter. The comb filter transfer function *H*(*f*) is given as follows:

∞ ∑ *k*=−∞

*H*(*f*) =

with N = 1, 63.

### *3.2.3. Signal-to-noise ratio improvement*

The SNR of received signals can be significantly improved after the coherent combination in the comb filter, since the signal power grows quadratically, while the noise power grows only linearly. We first consider ideal components where there is no distortion in delay line and the comb filter has no loop gain. For the investigation of the SNR improvement, an AWGN channel is sufficient. A train of modulated UWB impulses and additive noise is the input signal to the multiplier. The SNR at the input of the multiplier is compared to the SNR after the comb filter to calculate the comb filter processing gain *Gp*. Fig. 27(a) shows an example of an input signal with a SNR = -15 dB (upper) compared to the output signal (lower). The SNR improvement can be seen clearly as the receiving impulse becomes visible after a few iterations. This comb filter signal processing results in a SNR improvement of 10 log(*N*) dB, where *N* is the number of iterations. The UWB signal is algebraically added, therefore the signal energy is increased by a factor of N2. On the other hand, the noise contributions in each chip are added in power, and therefore the noise energy within the symbol interval is increased only by a factor of N. For communications, the BER of PPM with correlation and energy detection also show the same improvement. The BER performance for DS-PPM with *N* = 1 and *N* = 63 using M sequence is shown in Fig. 27(b) . Using several chips per symbol with the comb filter approach gives a performance gain with respect to SNR but not to *Eb*/*N*<sup>0</sup> due to the difference in the data rate. For UWB transmission, the SNR cannot be improved by increasing the transmitted energy because of the spectral mask limitation. A trade-off between data rate and the SNR improvement at the receiving side has to be made.

**Figure 27.** (a) Input and output of the comb filter with remodulation. (b) BER performance of DS-PPM with N = 1, 63.

For physical systems this calculation is only valid if no instability is introduced. Therefore the values of the gain *Gc* of the comb filter can only be in the range of 0-1, because otherwise an oscillation occurs. *Gc* is designed to be one but considering real components, a loss in the loop is possible. The comb filter processing gain *Gp* is a function of the gain *Gc* of the comb filter and the number of iterations *N*. It can be calculated as

$$G\_p = 10 \cdot \log\_{10} \left( \frac{\left( \sum\_{n=0}^{N-1} G\_c^n \right)^2}{\sum\_{n=0}^{N-1} G\_c^{2n}} \right) \tag{7}$$

The maximum processing gain *Gp* of the comb filter is 10 log(*N*) dB, when *Gc* = 1. The processing gain is reduced if *Gc* is less than one. The relations between *Gp*, *Gc* and *N* are shown in Fig. 28. We can see from Fig. 28(a) that the processing gain *Gp* saturated in the case where *Gc* is less than one because the impulse energy vanishes after some iterations. With a higher number of impulses per symbol, the comb filter loop gain has to be controlled more precisely as shown in Fig. 28(b).

### *3.2.4. Narrowband interference*

22 Will-be-set-by-IN-TECH

We assume synchronization in a certain time window between different users. With this assumption, we mainly consider m-sequences because they have low crosscorrelation and are simple to generate. Alternatively, binary zero-correlation-zone (ZCZ) sequence can also be considered. Unlike m-sequences, these sequences have a crosscorrelation of zero within a small window but the drawback is a smaller number of sequence in one set [40]. Binary ZCZ sequence sets were investigated in [37] for communication because of their potential to completely eliminate multi-sensor and interchip interference (ICI). A near-far problem was

The SNR of received signals can be significantly improved after the coherent combination in the comb filter, since the signal power grows quadratically, while the noise power grows only linearly. We first consider ideal components where there is no distortion in delay line and the comb filter has no loop gain. For the investigation of the SNR improvement, an AWGN channel is sufficient. A train of modulated UWB impulses and additive noise is the input signal to the multiplier. The SNR at the input of the multiplier is compared to the SNR after the comb filter to calculate the comb filter processing gain *Gp*. Fig. 27(a) shows an example of an input signal with a SNR = -15 dB (upper) compared to the output signal (lower). The SNR improvement can be seen clearly as the receiving impulse becomes visible after a few iterations. This comb filter signal processing results in a SNR improvement of 10 log(*N*) dB, where *N* is the number of iterations. The UWB signal is algebraically added, therefore the signal energy is increased by a factor of N2. On the other hand, the noise contributions in each chip are added in power, and therefore the noise energy within the symbol interval is increased only by a factor of N. For communications, the BER of PPM with correlation and energy detection also show the same improvement. The BER performance for DS-PPM with *N* = 1 and *N* = 63 using M sequence is shown in Fig. 27(b) . Using several chips per symbol with the comb filter approach gives a performance gain with respect to SNR but not to *Eb*/*N*<sup>0</sup> due to the difference in the data rate. For UWB transmission, the SNR cannot be improved by increasing the transmitted energy because of the spectral mask limitation. A trade-off between

**Figure 26.** Illustration of multiuser interference suppression at the comb filter.

taken into account and we can see an advantage of this sequence set.

data rate and the SNR improvement at the receiving side has to be made.

*3.2.3. Signal-to-noise ratio improvement*

Since UWB covers a very large bandwidth, strong interference within the band is possible and can cause problems at the receiver. Comb filter in combination with multiplication with the spreading waveform can suppress narrowband interference very well. Only a periodic signal that has a period which equals one or multiples of the comb filter delay can go through the comb filter. The comb filter transfer function *H*(*f*) is given as follows:

$$H(f) = \sum\_{k=-\infty}^{\infty} T\_s \cdot \text{sinc}(T\_s(f - k/T\_c)) \exp(-j\pi f T\_s) \tag{8}$$

We can see that the transfer function of the comb filter consists of several peaks. The peaks could be seen as tunnels that allow only signals with specific frequencies to pass. After the

**Figure 28.** Relationship between processing gain *Gp*, comb filter loop gain *Gc* and impulses per symbol N.

(a) (b)

463

UWB in Medicine – High Performance UWB Systems for Biomedical Diagnostics and Short Range Communications

of the receiver for this concept is shown in Fig. 30. Note that the spreading sequences for the rake branches are cyclic-shifted versions of one single sequence. For DS-CSK, two parallel branches with different corresponding spreading sequence sets are needed. The results from

In comparison to our basic concept described before, the delay is shorter. If the delay is shortened, for example, from 12 ns to 2 ns, this is much more realistic with regard to the realization. The performance of the new concept was verified by simulations. We used UWB impulses within a band from 4 GHz to 6 GHz. Note that this band is due to the Wireless Body Area Networks (WBANs) channel model we used [12] but it is not important for our concept, it could be any. The channel model consists of two groups of paths with a lognormal fading statistic for each group. We used 4000 channel realizations. The total propagation time of the channels is 12 ns. Eight users/sensors with the same average receive energy are considered. The spreading sequences were m-sequences with *N* = 127. The receive signal at the input of the comb filter is shown in the upper part of Fig. 31(a). Additive noise is not considered here to focus only on the ICI and MUI suppression. We can see that both, ICI and MUI are very strong here. The SIR in this example is -17 dB. In the lower part of Fig. 31(a) the original channel impulse response (black) is compared to the results taken from different rake branches (red).

**Figure 29.** (a) Input and output of the comb filter with remodulation. (b) BER performance with

**Figure 30.** Block diagram of a rake-like comb filter receiver with energy detection.

From the results we can see that the interference is suppressed very well.

SIR = -15 dB.

two branches are then compared.

multiplication with the spreading waveform in the receiver, the spectrum of the UWB signal has a shape matched to the transfer function of the comb filter (signal with period *Tc*). On the other hand, the narrowband interferer is spread by this multiplication and only some frequency components could go through the comb filter. The width of each peak depends on the number of chips per symbol. It gets smaller as the number of summation steps in the comb filter increases and as a result more interference is suppressed.

The improvement of the signal-to-interference ratio (SIR) is demonstrated in Fig. 29(a). The signals at the input of the comb filter are the UWB signal and a narrowband interference consisting of an IEEE 802.11a OFDM WLAN signal with a bandwidth of 16.66 MHz and a center frequency of 5.2 GHz. In this example, the input SIR of -15 dB is improved by 10 dB for *N* = 63. In addition, the BER performance of DS-OOK and DS-PPM in an AWGN channel with the same interference is shown in Fig. 29(b). The performance for both methods is improved with increasing number of chips per symbol. The degradation of the performance due to the narrowband interference for DS-PPM is much less than that for DS-OOK. The narrowband interferer gives a contribution to both integrator outputs for DS-PPM, and because the outputs are compared, the influence is reduced. For the DS-OOK the influence remains.

### *3.2.5. Shortened delay comb filter*

The main challenge for implementing the comb filter based receiver is to realize the true wideband analog delay element. Shortening the delay means that overlapping channel impulse responses can occur at the receiver, the channel impulse response becomes longer than the chip interval *Tc*. As a result, after the comb filtering we get an amplified window-cut-out of the true impulse response with the window width *Tc*. It is shown in [35] that we can control the position of this window by adjusting the spreading sequence at the remodulation. The property of being able to extract different parts of the channel impulse response gives an opportunity to construct a receiver by using a rake concept. This means that for each part of the impulse response we have one rake branch where we calculate the energy. The modulation technique that is well-suited for this structure is DS-CSK. The block diagram

24 Will-be-set-by-IN-TECH

(a) (b)

**Figure 28.** Relationship between processing gain *Gp*, comb filter loop gain *Gc* and impulses per symbol

multiplication with the spreading waveform in the receiver, the spectrum of the UWB signal has a shape matched to the transfer function of the comb filter (signal with period *Tc*). On the other hand, the narrowband interferer is spread by this multiplication and only some frequency components could go through the comb filter. The width of each peak depends on the number of chips per symbol. It gets smaller as the number of summation steps in the comb

The improvement of the signal-to-interference ratio (SIR) is demonstrated in Fig. 29(a). The signals at the input of the comb filter are the UWB signal and a narrowband interference consisting of an IEEE 802.11a OFDM WLAN signal with a bandwidth of 16.66 MHz and a center frequency of 5.2 GHz. In this example, the input SIR of -15 dB is improved by 10 dB for *N* = 63. In addition, the BER performance of DS-OOK and DS-PPM in an AWGN channel with the same interference is shown in Fig. 29(b). The performance for both methods is improved with increasing number of chips per symbol. The degradation of the performance due to the narrowband interference for DS-PPM is much less than that for DS-OOK. The narrowband interferer gives a contribution to both integrator outputs for DS-PPM, and because the outputs

The main challenge for implementing the comb filter based receiver is to realize the true wideband analog delay element. Shortening the delay means that overlapping channel impulse responses can occur at the receiver, the channel impulse response becomes longer than the chip interval *Tc*. As a result, after the comb filtering we get an amplified window-cut-out of the true impulse response with the window width *Tc*. It is shown in [35] that we can control the position of this window by adjusting the spreading sequence at the remodulation. The property of being able to extract different parts of the channel impulse response gives an opportunity to construct a receiver by using a rake concept. This means that for each part of the impulse response we have one rake branch where we calculate the energy. The modulation technique that is well-suited for this structure is DS-CSK. The block diagram

are compared, the influence is reduced. For the DS-OOK the influence remains.

filter increases and as a result more interference is suppressed.

*3.2.5. Shortened delay comb filter*

N.

**Figure 29.** (a) Input and output of the comb filter with remodulation. (b) BER performance with SIR = -15 dB.

of the receiver for this concept is shown in Fig. 30. Note that the spreading sequences for the rake branches are cyclic-shifted versions of one single sequence. For DS-CSK, two parallel branches with different corresponding spreading sequence sets are needed. The results from two branches are then compared.

**Figure 30.** Block diagram of a rake-like comb filter receiver with energy detection.

In comparison to our basic concept described before, the delay is shorter. If the delay is shortened, for example, from 12 ns to 2 ns, this is much more realistic with regard to the realization. The performance of the new concept was verified by simulations. We used UWB impulses within a band from 4 GHz to 6 GHz. Note that this band is due to the Wireless Body Area Networks (WBANs) channel model we used [12] but it is not important for our concept, it could be any. The channel model consists of two groups of paths with a lognormal fading statistic for each group. We used 4000 channel realizations. The total propagation time of the channels is 12 ns. Eight users/sensors with the same average receive energy are considered. The spreading sequences were m-sequences with *N* = 127. The receive signal at the input of the comb filter is shown in the upper part of Fig. 31(a). Additive noise is not considered here to focus only on the ICI and MUI suppression. We can see that both, ICI and MUI are very strong here. The SIR in this example is -17 dB. In the lower part of Fig. 31(a) the original channel impulse response (black) is compared to the results taken from different rake branches (red). From the results we can see that the interference is suppressed very well.

#### 26 Will-be-set-by-IN-TECH 464 Ultra-Wideband Radio Technologies for Communications, Localization and Sensor Applications UWB in Medicine – High Performance UWB Systems for Biomedical Diagnostics and Short Range Communications <sup>27</sup>

We also look at the BER performance of this new concept in Fig. 31. A WBAN channel is used. For our original basic comb filter concept, *Tc* = 12 ns and 6 ns are considered. We label them as 'Basic 1' and 'Basic 2', respectively. For the rake-like receiver, we consider *Tc* = 2 ns with 6 rake branches which yield equivalent integration time of 12 ns. We can see that the performance of our original concept is getting worse if the delay is shortened. There is a loss of about 4 dB at the BER for 10−4. The result for 'Rake-like' is comparable to 'Basic 1'. The results show that we can achieve a very similar performance to the original concept with much shorter analog delay elements.

Prediction:

Update:

estimation.

after the correlator.

*<sup>p</sup>*(x*k*|z1:*k*−1) =

details about the setup can be found in Sec. 4.1.

*<sup>p</sup>*(x*k*|z1:*k*) = *<sup>p</sup>*(z*k*|x*k*)*p*(x*k*|z1:*k*−1)

We can see that the likelihood function *<sup>p</sup>*(z*k*|x*k*) and the movement model *<sup>p</sup>*(x*k*|x*k*−1) are important items in Bayesian estimation. Unfortunately the posterior density is usually intractable but there are several ways to implement the algorithm. Particle filtering deals with this problem by using samples (particles) with associated weights to represent the posterior density. If the number of particles is sufficiently large, the estimates reach optimal Bayesian

A setup consisting of a bistatic UWB transceiver and a moving metal plate is used for demonstrating the use of particle filtering in IR-UWB tracking. A 5th derivative Gaussian impulse fitting to the FCC mask is used and the repetition rate is 200 MHz (i.e. *Ts* = 5 ns). The corresponding distance is 75 cm. The metal plate moves periodically within 5 mm range. A schematic block diagram of the setup is shown in Fig. 32(a). The receive signal consists of two main contributions which can be seen in each period, corresponding to two-path propagation on the channel. The first path is the direct path between the transmit and the receive antennas (direct coupling signal). This signal has very strong visible ringing due to signal reflection from the impedance mismatch. The second signal is a signal reflected from the moving metal plate. An example of a receiving signal with a target 30 cm away is shown in Fig. 32(b). More

(a) Measurement setup (b) Receive signal

**Figure 32.** Measurement setup for target movement tracking and example of a periodic receive signal

Particle filtering with a 2-path model is used for tracking the transmission delay of the two paths. The signal after the impulse correlation can be represented in discrete-time in each

*<sup>p</sup>*(x*k*|x*k*−1)*p*(x*k*−1|z*k*−1)*d*x*k*−<sup>1</sup> (9)

*<sup>p</sup>*(z1:*k*|z1:*k*−1) (10)

UWB in Medicine – High Performance UWB Systems for Biomedical Diagnostics and Short Range Communications

465

**Figure 31.** (a) Input signal of the comb filter (upper). Comparison between the actual channel impulse response and the output signal of the rake branches (lower). (b) Comparison of the bit error rate from different receivers.

### **3.3. Particle filtering**

In this part we want to address an algorithm for movement tracking which can be used for radar and imaging application. The background is the application of IR-UWB for biomedical diagnostics, e.g. vital sign detection (like breathing/heart rate) and also tracking of body movements to compensate for errors which degrade the quality of inner body imaging. Simple methods such as tracking the maximum/minimum of the receive signal in a specific area perform reasonably well. A problem occurs if the peak of the signal cannot be easily identified. In cases with high attenuation, the estimation can become very bad for those simple methods. Another problem is caused by stationary echoes (clutter). Particle filtering can help in this situation. It is a technique which implements a sequential Bayesian estimation by using Monte-Carlo methods. Particle filtering is commonly known to be used in localization applications and tracking in dynamic scenarios [18]. The estimation uses a movement model to incorporate the temporal correlation of the change of unknown parameters.

The Bayesian estimator finds unknown parameters x*<sup>k</sup>* (signal delays) from a set of measurement signals z1:*<sup>k</sup>* using a posterior density function *p*(x*k*|z1:*k*). It is very common that the algorithm is processed in a recursive manner. In summary, the algorithm consists of a 'prediction' phase and an 'update' phase, where the prior density *<sup>p</sup>*(x*k*|z1:*k*−1) and the posterior density *p*(x*k*|z1:*k*) are estimated every time step *k*.

Prediction:

26 Will-be-set-by-IN-TECH

We also look at the BER performance of this new concept in Fig. 31. A WBAN channel is used. For our original basic comb filter concept, *Tc* = 12 ns and 6 ns are considered. We label them as 'Basic 1' and 'Basic 2', respectively. For the rake-like receiver, we consider *Tc* = 2 ns with 6 rake branches which yield equivalent integration time of 12 ns. We can see that the performance of our original concept is getting worse if the delay is shortened. There is a loss of about 4 dB at the BER for 10−4. The result for 'Rake-like' is comparable to 'Basic 1'. The results show that we can achieve a very similar performance to the original concept with much shorter analog

(a) (b)

**Figure 31.** (a) Input signal of the comb filter (upper). Comparison between the actual channel impulse response and the output signal of the rake branches (lower). (b) Comparison of the bit error rate from

In this part we want to address an algorithm for movement tracking which can be used for radar and imaging application. The background is the application of IR-UWB for biomedical diagnostics, e.g. vital sign detection (like breathing/heart rate) and also tracking of body movements to compensate for errors which degrade the quality of inner body imaging. Simple methods such as tracking the maximum/minimum of the receive signal in a specific area perform reasonably well. A problem occurs if the peak of the signal cannot be easily identified. In cases with high attenuation, the estimation can become very bad for those simple methods. Another problem is caused by stationary echoes (clutter). Particle filtering can help in this situation. It is a technique which implements a sequential Bayesian estimation by using Monte-Carlo methods. Particle filtering is commonly known to be used in localization applications and tracking in dynamic scenarios [18]. The estimation uses a movement model

The Bayesian estimator finds unknown parameters x*<sup>k</sup>* (signal delays) from a set of measurement signals z1:*<sup>k</sup>* using a posterior density function *p*(x*k*|z1:*k*). It is very common that the algorithm is processed in a recursive manner. In summary, the algorithm consists of a 'prediction' phase and an 'update' phase, where the prior density *<sup>p</sup>*(x*k*|z1:*k*−1) and the

to incorporate the temporal correlation of the change of unknown parameters.

posterior density *p*(x*k*|z1:*k*) are estimated every time step *k*.

delay elements.

different receivers.

**3.3. Particle filtering**

$$p(x\_k|z\_{1:k-1}) = \int p(x\_k|x\_{k-1})p(x\_{k-1}|z\_{k-1})dx\_{k-1} \tag{9}$$

Update:

$$p(x\_k|z\_{1:k}) = \frac{p(z\_k|x\_k)p(x\_k|z\_{1:k-1})}{p(z\_{1:k}|z\_{1:k-1})} \tag{10}$$

We can see that the likelihood function *<sup>p</sup>*(z*k*|x*k*) and the movement model *<sup>p</sup>*(x*k*|x*k*−1) are important items in Bayesian estimation. Unfortunately the posterior density is usually intractable but there are several ways to implement the algorithm. Particle filtering deals with this problem by using samples (particles) with associated weights to represent the posterior density. If the number of particles is sufficiently large, the estimates reach optimal Bayesian estimation.

A setup consisting of a bistatic UWB transceiver and a moving metal plate is used for demonstrating the use of particle filtering in IR-UWB tracking. A 5th derivative Gaussian impulse fitting to the FCC mask is used and the repetition rate is 200 MHz (i.e. *Ts* = 5 ns). The corresponding distance is 75 cm. The metal plate moves periodically within 5 mm range. A schematic block diagram of the setup is shown in Fig. 32(a). The receive signal consists of two main contributions which can be seen in each period, corresponding to two-path propagation on the channel. The first path is the direct path between the transmit and the receive antennas (direct coupling signal). This signal has very strong visible ringing due to signal reflection from the impedance mismatch. The second signal is a signal reflected from the moving metal plate. An example of a receiving signal with a target 30 cm away is shown in Fig. 32(b). More details about the setup can be found in Sec. 4.1.

**Figure 32.** Measurement setup for target movement tracking and example of a periodic receive signal after the correlator.

Particle filtering with a 2-path model is used for tracking the transmission delay of the two paths. The signal after the impulse correlation can be represented in discrete-time in each

period and is considered as measurement signal *zk*. The delay of each paths form the state vector *xk*. More details on the setup can be found in [36].

In the following, tracking results for a moving metal plate with distances of approximately 90 cm and 120 cm are discussed. Since the distance of ambiguity for the measurement setup is 75 cm, we use the knowledge that the target is in the 75-150 cm range. This is sufficiently large for our target application. The reflected signal appears one period after the original impulse was transmitted. Particle filtering with 1000 particles is considered and the results are compared with a conventional maximum tracking method. The movement of the first path (direct coupling) and the second path (metal plate) are tracked simultaneously.

We first consider a setup with the moving target at approximately 120 cm. An example for one period of the receive signal is shown in Fig. 33(a). The reflected signal is located at around 8.8 ns. In this example, the reflected signal can be easily recognized. Fig. 33(b). shows the tracking results of the moving metal plate from both methods (particle filtering and maximum tracking for comparison). The tracking results fit well with each other and the small movement of 5 mm was estimated correctly. We can see that the particle filtering is more robust. This improvement comes from the fact that the movement model incorporates the temporal correlation of the change of the channel delays in different time steps. The conventional method does not use this information and the results can change rapidly. The particle filter needs some iterations to converge to the correct estimate.

(a) Receive signal (b) Tracking results

467

UWB in Medicine – High Performance UWB Systems for Biomedical Diagnostics and Short Range Communications

(a) Receive signal (b) Tracking results

Usually, the drawback of the particle filter is its complexity. In the applications considered here, this drawback is not so serious, because our system model is relatively simple. Using

**Figure 35.** Receive signal with target distance ≈ 90 cm and tracking result for target distance ≈ 90 cm

from conventional maximum tracking and particle filtering.

**Figure 34.** Receive signal with target distance ≈ 120 cm with additional noise and tracking result for target signal ≈ 120 cm with additional noise from conventional maximum tracking and particle filtering. Now we consider the tracking of the reflected signal for the moving metal plate at a distance of 90 cm. An example of a receive signal is shown in Fig. 35(a). We can see that the target signal is in the same interval as the ringing of the direct coupling signal. The amplitudes of the target signal and the ringing are comparable. It is not easy to distinguish between these two signals anymore. This is the same situation as in radar where we have cluttering. It can cause a bias to the estimation. A comparison of the tracking results from maximum tracking and particle filtering is shown in Fig. 35(b). We can see that the maximum tracking performs very badly because of the bias. The tracking results from particle filtering are very good. The use of the multipath propagation model eliminates the error bias given by the clutter (direct

coupling).

**Figure 33.** Receive signal with target distance ≈ 120 cm and tracking result for target distance ≈ 120 cm from conventional maximum tracking and particle filtering.

In this part, we consider the case where the noise is strong and the peak of the signal cannot be easily identified. To reduce the SNR of the signal, Gaussian noise was added to the measurement data of the previous part, so that the SNR was approximately 0dB. An example for one period of the receive signal *zk* is shown in Fig. 34(a). We can see that the target signal is not clearly visible anymore and the peak value is disturbed strongly by noise. The tracking results are shown in Fig. 34(b). The conventional method does not work. The particle filtering still gives good estimates, because it does not consider only the maximum point in the target signal but the waveform as a whole. The movement model also plays a role in this improvement.

28 Will-be-set-by-IN-TECH

period and is considered as measurement signal *zk*. The delay of each paths form the state

In the following, tracking results for a moving metal plate with distances of approximately 90 cm and 120 cm are discussed. Since the distance of ambiguity for the measurement setup is 75 cm, we use the knowledge that the target is in the 75-150 cm range. This is sufficiently large for our target application. The reflected signal appears one period after the original impulse was transmitted. Particle filtering with 1000 particles is considered and the results are compared with a conventional maximum tracking method. The movement of the first

We first consider a setup with the moving target at approximately 120 cm. An example for one period of the receive signal is shown in Fig. 33(a). The reflected signal is located at around 8.8 ns. In this example, the reflected signal can be easily recognized. Fig. 33(b). shows the tracking results of the moving metal plate from both methods (particle filtering and maximum tracking for comparison). The tracking results fit well with each other and the small movement of 5 mm was estimated correctly. We can see that the particle filtering is more robust. This improvement comes from the fact that the movement model incorporates the temporal correlation of the change of the channel delays in different time steps. The conventional method does not use this information and the results can change rapidly. The

(a) Receive signal (b) Tracking results

**Figure 33.** Receive signal with target distance ≈ 120 cm and tracking result for target distance ≈ 120 cm

In this part, we consider the case where the noise is strong and the peak of the signal cannot be easily identified. To reduce the SNR of the signal, Gaussian noise was added to the measurement data of the previous part, so that the SNR was approximately 0dB. An example for one period of the receive signal *zk* is shown in Fig. 34(a). We can see that the target signal is not clearly visible anymore and the peak value is disturbed strongly by noise. The tracking results are shown in Fig. 34(b). The conventional method does not work. The particle filtering still gives good estimates, because it does not consider only the maximum point in the target signal but the waveform as a whole. The movement model also plays a role in this

path (direct coupling) and the second path (metal plate) are tracked simultaneously.

particle filter needs some iterations to converge to the correct estimate.

from conventional maximum tracking and particle filtering.

improvement.

vector *xk*. More details on the setup can be found in [36].

**Figure 34.** Receive signal with target distance ≈ 120 cm with additional noise and tracking result for target signal ≈ 120 cm with additional noise from conventional maximum tracking and particle filtering.

Now we consider the tracking of the reflected signal for the moving metal plate at a distance of 90 cm. An example of a receive signal is shown in Fig. 35(a). We can see that the target signal is in the same interval as the ringing of the direct coupling signal. The amplitudes of the target signal and the ringing are comparable. It is not easy to distinguish between these two signals anymore. This is the same situation as in radar where we have cluttering. It can cause a bias to the estimation. A comparison of the tracking results from maximum tracking and particle filtering is shown in Fig. 35(b). We can see that the maximum tracking performs very badly because of the bias. The tracking results from particle filtering are very good. The use of the multipath propagation model eliminates the error bias given by the clutter (direct coupling).

**Figure 35.** Receive signal with target distance ≈ 90 cm and tracking result for target distance ≈ 90 cm from conventional maximum tracking and particle filtering.

Usually, the drawback of the particle filter is its complexity. In the applications considered here, this drawback is not so serious, because our system model is relatively simple. Using particle filtering in parallel with conventional methods and exchange information between both methods can also help to reduce complexity.

In the second step, the transmitter inside of the dielectric is switched on and the radar sensors operate in receive mode recording the time of flight of the transmitted signal. Finally, we

Several methods to estimate the surface of a highly reflective medium using UWB pulse radar sensors have been investigated in recent years [17, 29]. Some of these imaging algorithms, however, need extensive preprocessing of the measurement data or suffer from high complexity and computation time. In the first part of this section, we derive a simple and easy to implement 3D surface estimation algorithm based on trilateration. In the second part, building on this surface estimation method, we present an approach for the localization of transmitters inside an arbitrarily shaped dielectric medium taking into account its surface

Radar measurements with quasi-omnidirectional antennas only provide information about the target distance, but not about its direction. This makes surface imaging an inverse problem which can only be solved by combining measurement results from different antenna positions. In this context, target ranging using trilateration means determining the intersections of spheres, the radii of which correspond to measured target distances. The underlying assumption for using trilateration as a surface estimation method is that two neighboring antennas are "seeing" the same scattering center. As with other imaging algorithms this

The imaging principle shall first be explained using a two-dimensional example. Fig. 37(a) shows the measurement scenario of a linear array of monostatic radar transceivers arranged along the *x*-axis scanning the surface of a target in *z*-direction. Each array element measures the distance to the closest point on the target. Two exemplary measurement points *Xn* and *Xn*+<sup>1</sup> are picked out, and semi circles with radii corresponding to the measured target distances are plotted around the antennas. The estimated surface point is the intersection of

x

**Figure 37.** Cross section of a 2D imaging problem using a linear array of monostatic radar transceivers along the *x*-axis (a) and top view of a 3D imaging setup showing three measuring points of an antenna

Three-dimensional imaging demands for a third antenna position located in a different dimension. This setup is shown as a top view in Fig. 37(b). At each of the three positions

r

1

A

3

A

j

z

d x

y

469

r

3

r

2

UWB in Medicine – High Performance UWB Systems for Biomedical Diagnostics and Short Range Communications

(b) Top view of 3D imaging setup

A

1 2

analyze all the acquired data and determine the position of the transmitter.

*3.4.1. Surface estimation algorithm based on trilateration*

X X <sup>n</sup>

n (a) 2D imaging principle

+1

r r n n +1

intersection

assumption can lead to inaccuracies of estimated target points.

profile.

the two circles.

target

array in the *x*-*y*-plane (b).

z
