1686 IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, VOL. 25, NO.

9, SEPTEMBER 2014

Fading Channel Prediction Based on Combination

of Complex-Valued Neural Networks and
Chirp Z-Transform
Tianben Ding and Akira Hirose, Fellow, IEEE

Abstract— Channel prediction is an important process for time-domain (TD) and frequency-domain methods. However,
channel compensation in a fading environment. If a future each of them has problems more or less in its calculation cost
channel characteristic is predicted, adaptive techniques, such as or prediction precision or both. For instance, the accuracy of
pre-equalization and transmission power control, are applicable
before transmission in order to avoid degradation of commu- linear or nonlinear prediction in TD is not high in general
nications quality. Previously, we proposed channel prediction because the channel observed in the TD often yields irregular
methods employing the chirp z-transform (CZT) with a linear variation. On the other hand, frequency-domain methods, such
extrapolation as well as a Lagrange extrapolation of frequency- as ROOT-MUSIC and ESPRIT, require large calculation cost
domain parameters. This paper presents a highly accurate for estimating the channel.
method for predicting time-varying channels by combining a
multilayer complex-valued neural network (CVNN) with the To realize a low-cost and high-precision prediction, we
CZT. We demonstrate that the channel prediction accuracy of previously proposed the use of chirp z-transform (CZT) [15]
the proposed CVNN-based prediction is better than those of the in Jakes model [1] for high-accuracy extraction of frequency-
conventional prediction methods in a series of simulations and domain channel parameters [16]. We also demonstrated
experiments. that the performance is greatly improved by combining
Index Terms— Channel prediction, chirp z-transform the CZT with linear extrapolation [17], [18] or Lagrange
(CZT), complex-valued neural networks (CVNNs), fading, extrapolation [19]. However, in a series of practical exper-
frequency domain, high-capacity spatial-domain multiple iments, we found that these methods fail to predict future
access (HC-SDMA).
channels in some situations.
I. I NTRODUCTION In this paper, we propose new prediction methods based
on complex-valued neural networks (CVNNs) [20] to predict
P ERFORMANCE of mobile communications often suffers
from various fading phenomena. To reduce the adverse
effect, there are some techniques such as diversity reception
the channel characteristics of respective paths represented
by channel parameters estimated by the CZT. The CVNNs,
dealing with a complex-valued signal as an entity, is a powerful
and error-correction coding. Other countermeasures are pre-
framework in the signal processing area [21], [22]. It also
equalization and transmission power control. Such adaptive
brings a flexible processing to the prediction of channel
techniques require channel prediction for the channel changes
characteristics, which is also complex-valued. We assume a
in time [1], [2]. The prediction also leads to high-capacity
multilayer (ML) and a real-time-recurrent-learning (RTRL)
spatial-domain multiple access (HC-SDMA) [3] and multiuser
CVNN. Simulation results demonstrate that the CZT-combined
multiple-input multiple-output [4] systems.
ML-CVNN (CZT-ML-CVNN)-based prediction can predict
There exist several methods in channel prediction. We can
the channel with a very high accuracy, and keep this high
predict channel characteristic using, e.g., linear prediction
accuracy even in situations, where conventional predictions
based on linear extrapolation [5], [6] and autoregressive (AR)
fail. In particular, we find that the proposed ML-CVNN-based
model focusing on serial autocorrelation in time [7]–[10].
prediction presents a higher performance in the case of low
In addition, we may also employ recurrent neural network for
frequency-resolution transform, which is a necessary condition
nonlinear prediction [11], [12]. Besides, we can use super-
to restrain the calculation cost. Furthermore, a series of exper-
resolution methods with high-precision spectral estimation
iments also demonstrate that the proposed ML-CVNN-based
such as ROOT-MUSIC [13] and ESPRIT [14] to extrapolate
prediction is more accurate than the conventional predictions
channel characteristics. These methods can be categorized into
in actual fading channels.
Manuscript received January 30, 2013; revised January 30, 2014; accepted This paper is organized as follows. Section II describes
February 9, 2014. Date of publication March 4, 2014; date of current version the combination of frequency and time domains for channel
August 15, 2014.
The authors are with the Department of Electrical Engineering and prediction as well as conventional prediction methods. In
Information Systems, University of Tokyo, Tokyo 113-8656, Japan (e-mail: Section III, we propose the prediction methods based on
[email protected]; [email protected]). CVNNs. Then, Sections IV and V present their performance in
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. simulations and experiments, respectively. Finally, Section VI
Digital Object Identifier 10.1109/TNNLS.2014.2306420 provides the conclusions.

2162-237X © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Authorized licensed use limited to: Rohde & Schwarz. Downloaded on February 18,2020 at 13:00:00 UTC from IEEE Xplore. Restrictions apply.
DING AND HIROSE: FADING CHANNEL PREDICTION BASED ON COMBINATION OF CVNNs AND CZT 1687

II. C HANNEL P REDICTION W ITH C OMBINATION OF

F REQUENCY-D OMAIN AND T IME -D OMAIN T ECHNIQUES
A. Channel Model and CZT
According to the Jakes model, a fading channel can be
modeled as the summation of sinusoids at a receiver, which are
the multipath rays caused by scattering and reflection. Each
sinusoid can be characterized by a set of path parameters such
as amplitude am , Doppler frequency f m , and phase shift φm .
The channel characteristic c(t) as a function of time t is the
summation of M complex signal paths and expressed as

M
c(t) = am e j (2π f m t +φm ) . (1)
m=1

The Doppler frequency f m is given by

fc
fm = vcos(ϕ) (2)
c
where v and c are the mobile user’s speed and the speed of
light, respectively, fc is the carrier frequency, and ϕ is the
incident radio wave angle with respect to the motion of the
mobile user.
We can utilize channel prediction methods to reduce the
influence of fading. There are two types of methods for
channel prediction, which predict the future channel charac-
teristic from the past channel information: 1) TD method,
which predicts c(t) in TD directly and 2) frequency-domain
method, which passes the past channel characteristic to the fre-
quency domain to separate multiple paths, and then extrapolate
them.
Previously, we proposed the combination of the time and
frequency-domain treatments. As shown in Fig. 1, we apply
the Fourier transform to the past channel information (t ≤ 0)
to observe a Doppler frequency spectrum. The peaks represent
the Doppler-shift paths. We estimate the path parameters of
each path by finding the peak amplitude âm (t) and its Doppler Fig. 1. Conceptual illustrations representing (a) CZT to convert the past
shift frequency fˆm (t) in the Doppler spectrum, and the phase channel information into the frequency domain for the extraction of the
frequency-domain channel parameters, namely, (b) amplitude am and Doppler
shift φ̂m (t) in the phase spectrum presented in Fig. 1(c), frequency f m , and (c) phase shift φm , respectively.
corresponding to the Doppler frequency peak. Note that the
accuracy of the path estimation depends on the frequency
resolution. fˆm (−n/2), φ̂m (−n/2), are different from the parameters at the
In general, the Doppler shift frequencies fˆm (t) are much prediction period (t > 0). Thus, we have to predict the chan-
smaller than the carrier frequency. Therefore, we need to nel characteristic c̃(t) based on these estimated information
observe precisely the frequency spectrum in the region close in TD.
to the zero frequency. A conventional nonzero-padding-FFT
requires a long observation period. Even a zero-padding-FFT
increases the calculation cost. To overcome these issues, we B. Linear and Lagrange-Based Predictions
use the CZT to obtain the Doppler spectrum. The CZT realizes Fig. 1(a) is also a schematic diagram showing our previous
low calculation cost as well as a smooth frequency-domain methods. As previous works, we proposed a linear predic-
interpolation, resulting in a high estimation accuracy of the tion method and a prediction method based on Lagrange
peak frequencies fˆm (t) [15], [16]. extrapolation that predict the parameters, ãm (t), f˜m (t), φ̃m (t)
There is, however, a problem even in the CZT to obtain (t ≥ 0), to predict the channel characteristic c̃(t). The linear
a future channel characteristic (t > 0) from the past path prediction extrapolates ãm (t), f˜m (t), φ̃m (t) (t ≥ 0) using the
parameters. As shown in Fig. 1(a), we apply a CZT window parameters at t = −n/2 and −n, namely, âm (−n), fˆm (−n),
to obtain the Doppler spectrum. Since the CZT window φ̂m (−n) and âm (−n/2), fˆm (−n/2), φ̂m (−n/2) [17], [18]. The
function, such as Hann window, has length of n, extracts the Lagrange-based prediction extrapolates ãm (t), f˜m (t), φ̃m (t)
channel parameters with emphasis on the central part. There- (t ≥ 0) based on Lagrange extrapolation using the parameters
fore, the latest estimated path parameters, namely, âm (−n/2), at t = −n/2, −n and −3n/2 [19]. Then, according with the

Authorized licensed use limited to: Rohde & Schwarz. Downloaded on February 18,2020 at 13:00:00 UTC from IEEE Xplore. Restrictions apply.
1688 IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, VOL. 25, NO. 9, SEPTEMBER 2014

Fig. 2. Schematic diagram of the proposed channel prediction method based on the CZT and CVNNs prediction of frequency-domain channel characteristics
cm of dominant paths.

Jakes model, we obtained channel characteristics of respective with information in the complex domain is suitable for treating
paths as rotation in the complex plane. The channel characteristic has
˜
c̃m (t) = ãm e j (2π f m t +φ̃m ) . (3) not only amplitude information, but also phase information.
The dominant part of channel changes of respective paths
Summation of the multiple complex paths c̃m (t) gives the is rotation in the complex plane. Therefore, the CVNNs
channel characteristic c̃(t). can bring about both high flexibility and low generalization
In the process of phase prediction, a long interval of window error in the channel prediction. In addition, we can deal
center generates a large phase increase/decrease, which leads with channel characteristics of respective paths c̃m (t) =
to violation of the sampling theorem, and fails to predict φ̃m (t) ãm e j θ̃m (where θ̃m ≡ 2π f˜m t + φ̃m ) as complex entities in the
properly. To solve this problem, we set the windows with an CVNNs. In this way, the channel prediction becomes more
interval shorter than n/2 as shown in Fig. 1(a), and avoid the appropriate to predict time-varying channel than prediction of
increase/decrease of φm beyond ±π. separate channel parameters ãm , f˜m , and φ̃m independently.
These predictions present high performances in most cases, Fig. 2 represents the prediction method presented in this
and the Lagrange-based prediction is more accurate than the paper. In this method, we set the windows with a shorter
linear prediction in some numerical experiments. However, we interval again to avoid the increase/decrease of φm beyond ±π
found that these predictions cannot predict future channels in as we did in the linear and Lagrange-based predictions. We
some situations. For example, the accuracy of these predictions estimate ĉm (t) with the CZT at I points: t = −n/2 − I +
deteriorates when scatterers’ distance has particular values. 1, . . . , −n/2, and feed them to a CVNN to predict
In these cases, the linear and Lagrange-based predictions fail c̃m (−n/2 + 1). After that, we predict c̃m (−n/2 + 2) by
to improve the performance or, in several cases, degrade the using ĉm (t) estimated at t = −n/2 − I + 2, . . . , −n/2 and
communications quality. c̃m (−n/2 + 1). We iterate this process to obtain c̃m (t) (t ≥ 0).
Summation of the multiple paths c̃m (t) gives the characteristic
C. Prediction Method Based on AR Model c̃(t) (t ≥ 0). In this stage, we also know the actual ĉ(0).
There is a general prediction method based on an AR model. Then, we combine ĉ(0) ≡ â(0)e j θ̂(0) with c̃(0) ≡ ã(0)e j θ̃(0)
The AR-model-based method predicts respective characteristic as â(0)ã(t)e j (θ̃(t )+θ̂(0)−θ̃(0)) /ã(0). This is the finally predicted
c̃m (t) using past channel characteristics ĉm (t) presented by the channel.
CZT as

IAR
B. ML CVNNs
c̃m (t) = dk (t)ĉm (t − k) (4)
k In this paper, we propose two different CVNN-based pre-
where IAR is the number of past channel samples, and dk (t) diction methods. The first one is a CZT-ML-CVNN prediction.
is the AR coefficient. The AR coefficient dk (t) is updated Fig. 3 shows the construction of the ML CVNN used in
according to the steepest descent method, which minimizes this paper to predict c̃m (t) for each path. It has input termi-
the difference of predicted c̃m (t) and a teacher signal ĉm (t). nals, a hidden-neuron layer and an output-neuron layer. The
input terminals of the layered CVNN distribute input signals,
cm (t − 1), …, cm (t − IML ), to the hidden-layer neurons as
III. P ROPOSAL OF CVNN-BASED P REDICTION
their inputs z0 . In the same way, the outputs of the hidden-
A. Signals in CVNNs layer neurons z1 are passed to the output-layer neuron as its
In this paper, we propose channel prediction methods based inputs. The output of the output-layer neuron is the prediction
on combination of CVNNs and CZT. The CVNNs that deal result c̃m (t).

Authorized licensed use limited to: Rohde & Schwarz. Downloaded on February 18,2020 at 13:00:00 UTC from IEEE Xplore. Restrictions apply.
DING AND HIROSE: FADING CHANNEL PREDICTION BASED ON COMBINATION OF CVNNs AND CZT 1689

j i ≡ θl j −θ(l−1)i −θl j i and r is the number of learning

where θlrot
iteration, whereas K 1 and K 2 are learning constants. The
teacher signals ẑ l j are for the outputs z l j in layer l. They are
backpropagated through the layered network to yield ẑ (l−1)i
as
∗
ẑ (l−1)i = f ap (ẑ l∗j wl j i )
= tanh(|ẑ l∗j wl j i |)e( j (−θ̂l j )+θl j i ) . (9)
In other words, the teacher signals in the hidden layer ẑ1
are the signals obtained through the backpropagation of the
teacher signal ẑ 2 = ĉm (t) in (9).
In a general neural network, connection weights are updated
until theirs convergence. However, such update iterations
Fig. 3. Layered CVNN. require a large learning cost that hinders a real-time appli-
cation. To solve this problem, we utilize the weak regularity
of the respective channel characteristics. That is, we stop
The connection weight wl j i to the i th input at the j th the update of the weights at a certain number of iteration
neuron in the layer l is expressed by its amplitude |wl j i | and regardless of the convergence for ĉm (t − 1). Then, the weights
phase θl j i . The internal state of the neuron u l j is obtained as are used as the initial values in the weights update for ĉm (t).
the summation of the inputs weighted by wl j i , particularly By repeating this learning continuously, we can reduce the
 
ul j ≡ wl j i z (l−1)i = |wl j i ||z (l−1)i |e j (θl j i +θ(l−1)i ) . (5) learning cost and follow the respective characteristics, which
i i
vary continuously.
The output z l j is given by adopting an amplitude-phase-type
activation function f ap to u l j as C. Recurrent CVNNs Based on RTRL
Recurrent neural networks (RNNs) is widely used in time-
z l j = f ap (u l j ) = tanh(|u l j |)e( j arg(ul j )) . (6) domain prediction. A RNN learns channel characteristic with
We update the connection weights in the ML-CVNN as recurrent output signals. The RTRL [25] requires only a single
follows. The past channel characteristics for t ≤ −n/2 weight update in unit time. We combine the CZT with a
of respective paths ĉm (t) are estimated from the Doppler recurrent CVNN updated by the RTRL to predict the respective
spectrum observed with the CZT. The time t represents the characteristics (CZT-RTRL-CVNN-based prediction).
window center of the CZT. The ML-CVNN regards this ĉm (t)
as the teacher signal and the previous channel characteristics IV. N UMERICAL E XPERIMENTS
ĉm (t − 1), . . . , ĉm (t − IML ) as the input signals. The output A. Simulation Setup
of the ML-CVNN z 2 is used to update the weights. Then, we
The performance of the proposed CVNN-based predic-
predict the channel characteristics of respective paths c̃m (t)
tions is discussed in this section. We evaluate the accu-
(t ≥ 0) with the connection weights resulting at t = −n/2.
racy of proposed methods in numerical experiments. We
To update the weights in this paper, we applied the steepest
apply the conventional methods for comparison, e.g., two
descent method which updates weights to minimize the dif-
TD methods, namely, the TD-Linear and TD-RTRL-CVNN-
ference El ≡ |zl − ẑl |2 /2, where zl are the outputs signal
based prediction and the methods combined with CZT to
and ẑl are the teacher signal in layer l. Thus, the weight
predict respective-path characteristics, namely, CZT-Linear,
amplitude |wl j i | and the phase θl j i are renewed as [23], [24]
CZT-Lagrange-based, and CZT-AR-model-based prediction as
∂ El well as nonprediction, which regards estimated âm (−n/2),
|wl j i |(r + 1)=|wl j i |(r ) − K 1
∂(|wl j i |) fˆm (−n/2), φ̂m (−n/2) as ãm (t), f˜m (t), φ̃m (t) (t ≥ 0) without

=|wl j i |(r ) − K 1 (1 − |z l j |2 ) prediction (CZT only).
The geometrical setup is shown in Fig. 4 and the simulation
×(|z l j | − |ẑ l j |cos(θl j − θ̂l j ))|z (l−1)i |cosθlrot parameters are shown in Table I. There are two scatterers
 ji separate by x. The mobile user moves in the direction of
|z (l−1)i |
−|z l j ||ẑ l j |sin(θl j − θ̂l j ) sinθlrot
ji (7) the arrow at v = 14 m/s. This is a hard situation for channel
|u l j |
prediction because of the fading channel varying very fast.
1 ∂ El
θl j i (r + 1)= θl j i (r ) − K 2 The base station (BS) and the mobile user have the line
|wl j i | ∂θl j i
 of sight. In urban area, the number of dominant multiple
= θl j i (r ) − K 2 (1 − |z l j |2 ) paths are generally about three paths. The carrier frequency
is f c = 2 GHz.
×(|z l j | − |ẑ l j |cos(θl j − θ̂l j ))|z (l−1)i |sinθlrot To compare the effectiveness of the channel prediction
 ji
|z (l−1)i | methods in actual application, we assume an orthogonal
+|z l j ||ẑ l j |sin(θl j − θ̂l j ) cosθlrot
ji (8)
|u l j | frequency-division multiplexing (OFDM) system. The OFDM

Authorized licensed use limited to: Rohde & Schwarz. Downloaded on February 18,2020 at 13:00:00 UTC from IEEE Xplore. Restrictions apply.
1690 IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, VOL. 25, NO. 9, SEPTEMBER 2014

TABLE II
OFDM AND C HANNEL -P REDICTION PARAMETERS

Fig. 4. Geometrical setup used in the simulation. There are two scatterers,
a BS and a mobile user in the simulation field. The BS and the mobile user
have the line of sight.

TABLE I
S IMULATION PARAMETERS

Fig. 6. General TDD frame format.

B. Prediction Accuracy in Low-Resolution Conditions

Table III shows the calculation costs of conventional and
proposed methods in which S is the symbol number in a CZT
window as listed in Table II. The CZT-combined methods need
O(S log S) more calculation cost for CZT than TD methods.
The second terms for the CZT-combined methods express
prediction costs in the respective methods. The proposed
CZT-ML-CVNN-based prediction needs largest calculation
cost for the prediction. However, the calculation cost of
the second term is sufficiently small than that of the first
term even in the CZT-ML-CVNN-based prediction because
always S log S  M·RML ·JML ·IML for any practical window
Fig. 5. Block diagram of applied OFDM system. length n. The inequality shows that the window lengths in
the CZT is the parameter most influential to the calculation
systems can compensate frequency selective fading by itself, cost in the CZT-combined methods, including our proposed
but it requires some compensation for time-varying fading. The methods. A short window length is required for reducing the
block diagram of the OFDM is shown in Fig. 5. Table II lists calculation cost, but it results in a low frequency resolution.
the system parameters following the standard of HC-SDMA as In such a tradeoff situation, we have to predict channel
well as the prediction parameters employed for the CZT-AR- characteristics c̃m (t) with low-frequency-resolution estima-
model-based, the CZT-ML-CVNN-based, and the both RTRL- tions ĉm (t). Among various cases, we present a set of pre-
CVNN-based predictions. In this simulation, we assume a diction results when the window length is short. The length
QPSK as the modulation method and a time division duplex of Hann window is 8 TDD frame. Fig. 7 shows a result
(TDD), as shown in Fig. 6, which use a single frequency of the scatterers’ distance is x = 34 m. Fig. 7(a) shows
band for uplink and downlink in a time division manner. The the predicted relative power in the prediction frame (5 ms),
number of OFDM subcarriers is 52. We predict the change whereas Fig. 7(b) is the predicted phase. The results of the
of the channel characteristic for a unit TDD frame (5 ms) CZT-ML-CVNN-based prediction are near to the actual signal
from t = 0. The center of windows are set with an interval of while those of other methods case deviate from. Fig. 8 shows
1 frame. The prediction parameters were chosen in simulation the bit-error rate (BER) versus the signal-to-noise ratio in
in such a way that the prediction error is minimized. the OFDM system with channel compensations using the

Authorized licensed use limited to: Rohde & Schwarz. Downloaded on February 18,2020 at 13:00:00 UTC from IEEE Xplore. Restrictions apply.
DING AND HIROSE: FADING CHANNEL PREDICTION BASED ON COMBINATION OF CVNNs AND CZT 1691

TABLE III
C ALCULATION C OSTS OF T ESTED M ETHODS
the conventional methods degrade much more. In particular,
the performance of the CZT-AR-model, CZT-RTRL-CVNN,
and TD-Linear are worse by 7–10 dB. The other methods
completely fail to compensate the fading. The results show
that the prediction based on the CZT-ML-CVNN has a higher
performance even in the situation in which other predictions
fail to predict the fading channel.
In this case, the RTRL-CVNN shows medium BER perfor-
mance. Basically, a recurrent structure realizes simple influ-
ence of input signals on the output with first-order decay. It is
sometimes difficult for the RTRL-CVNN to predict a channel
in rapidly changing environment. However, the CZT-RTRL-
CVNN shows a result better than that of the TD-RTRL-CVNN.

C. Performance Dependence on Window Length

In this section, we analyze a relationship of the window
length of CZT and the prediction performances of each
method. In phase shift keying, the prediction result of phase
values is more influential than that of relative power to
the communications quality. We present the results of phase
prediction in respective prediction methods here. Fig. 9 shows
prediction-phase error (PPE) in the predicted phase accu-
mulated in a unit frame when the Hann window length n
and the scatterers’ distance x are changed for n = 6–50
and x = 2–50 m. Fig. 9(a) shows the error averaged
for the all distance as function of n. Fig. 9(b) presents the
maximum error observed when the distance is changed. In
addition, in Table IV, we list the averaged and max error
Fig. 7. Predicted (a) relative power and (b) phase curves of the channel as
functions of time in the prediction period (t > 0).
at relatively short window, n = 8, with the error in the TD
methods for comparison. In Fig. 9 and Table IV, a smaller
phase error directly means a higher performance. In quaternary
phase shift keying (QPSK), the prediction failure is caused
by the PPE beyond π/8 at a symbol. In this paper, a unit
TDD frame consists of s = 2500 symbol according to the
HC-SDMA standard. Thus, the prediction fails if the phase
error shown in Fig. 9 go over ∼1000 rad·symbol. This
is a rough estimation. In an actual system, there exists
noise. In an observation of the phase error values and
the BER, we found that 500-rad·symbol error corresponds
to 2–4 dB degradation in the BER, 1000 rad·symbol to
10–15 dB, and 1500 rad·symbol results in complete failure
of communications.
In many applications, the length should be sufficiently short
for the reduction of the calculation cost. In the region of n ≤ 15
in Fig. 9(a), we find that the error of the CZT-ML-CVNN is
the lowest among the tested methods. At the same time, we
find that the error curves rise up in most cases.
The peak phase error curves in Fig. 9(b) are somewhat
similar to the corresponding curves in Fig. 9(a). However,
Fig. 8. BER curves obtained for TD-Linear prediction, TD-RTRL-CVNN-
based prediction, CZT only, CZT-Linear prediction, CZT-Lagrange-based
the CZT-ML-CVNN-based prediction result shows the lowest
prediction, CZT-AR-model-based prediction, CZT-RTRL-CVNN-based pre- error over a wider window-length region. Simultaneously, the
diction, CZT-ML-CVNN-based prediction, and actual channel characteristics peak error value in Fig. 9(b) is almost the same amount as the
when the geometry causes severe change in the Doppler effects.
mean error value in Fig. 9(a). These facts show that there exists
no serious degradation and that the stability of the CZT-ML-
prediction results. The BER with the compensation based on CVNN-based prediction is very high. Again, the performance
the CZT-ML-CVNN is worse only by 2 to 4 dB than that based of the CZT-ML-CVNN-based method is higher than those of
on the actual channel characteristic, whereas those based on other methods.

Authorized licensed use limited to: Rohde & Schwarz. Downloaded on February 18,2020 at 13:00:00 UTC from IEEE Xplore. Restrictions apply.
1692 IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, VOL. 25, NO. 9, SEPTEMBER 2014

Fig. 10. PPEs of CZT-ML-CVNN-based prediction for various n versus

iteration number of weight updates.

Fig. 11. PPE of CZT-ML-CVNN-based prediction versus ratio of hidden

neuron number to input terminal number.
Fig. 9. (a) Mean and (b) peak phase errors for various prediction methods,
namely, CZT only, CZT-Linear prediction, CZT-Lagrange-based prediction,
CZT-AR-model-based prediction, CZT-RTRL-CVNN-based prediction, and
CZT-ML-CVNN-based prediction. error is averaged and plotted versus the iteration number.
The PPE decreases with the increase of iteration number,
TABLE IV
and is less than or ∼500 rad·symbol when the iteration
M EAN AND P EAK PPE OF THE CZT-C OMBINED M ETHODS AT n = 8 AND
number is over 3. Thus, the proposed method succeeds to
THE T IME -D OMAIN M ETHODS
realize high prediction accuracy in a relatively small iteration
number.
To stop update at specific number is useful to reducing
learning cost. Learning cost affects prediction performance in
real-time applications. The learning cost of the RTRL is small
since the weights are updated only once for a single set of input
signals. On the other hand, the CZT-ML-CVNN phase error
for the case without iteration (RML = 1) takes a larger error
than the CZT-RTRL-CVNN phase error (709.5 rad·symbol).
However, the CZT-ML-CVNN-based prediction achieves a
phase error less than 500 rad·symbol by increasing the iteration
number only to 7 even in low frequency resolution (n = 8).
These facts show that the proposed CZT-ML-CVNN-based
D. Influence of Learning Iteration Number and Hidden
method has an ability to predict the channel with a higher
Neuron Number to Input Terminal Number Ratio
accuracy and a small learning cost.
In our proposed method, iteration number of ML-CVNN- The ML-CVNN input terminal number IML and the hid-
weight update is a parameter influential to prediction accuracy den neural number JML are also influential parameters. The
and cost. The influence to the prediction accuracy is shown influence to the prediction accuracy is shown in Fig. 11. We
in Fig. 10. The window length n is 8, 20, 30, 40 and the evaluate the influence of the two parameters relatively by
scatterers’ distance x is changed for x = 2–50 m. Each fixing the ML-CVNN input terminal number IML and chang-
situation is tested 25 times, and then the accumulated phase ing the ML-CVNN hidden neural number JML . The window

Authorized licensed use limited to: Rohde & Schwarz. Downloaded on February 18,2020 at 13:00:00 UTC from IEEE Xplore. Restrictions apply.
DING AND HIROSE: FADING CHANNEL PREDICTION BASED ON COMBINATION OF CVNNs AND CZT 1693

Base Station (BS)

50m Mobile User (MU)

Fig. 12. Locations of the mobile user and the BS, including the movement
direction (Google Maps, modified).

TABLE V
E XPERIMENT PARAMETERS Fig. 14. Actual (a) relative power and (b) phase without modulation received
at the BS in Fig. 12.

velocity of 12.5 m/s. The block diagram of the transmitter

and the receiver is shown in Fig. 13. The MS transmits
1.297-GHz nonmodulation wave from a monopole antenna,
whereas the BS receives the wave at another monopole
antenna. The received channel signal is mixed with 1.287-GHz
local oscillator wave after amplified with a low-noise amplifier,
and then extracted as an intermediate frequency signal of
10 MHz. After passing it to an amplifier and a band-pass filter
Fig. 13. Block diagram of the system used in the actual experiment. with 2-MHz bandwidth, we sample the channel information by
60 MHz. For the compatibility with the HC-SDMA standard,
length n is 8 and the scatterers’s distance x is changed we thin down the time-sequential data at 500 kHz (downsam-
for x = 2–50 m. Each situation for a certain set of x, pling). Fig. 14 presents an example of the time-sequential data
IML , and JML is tested 25 times, and then the accumulated showing relative power and phase observed at the BS. In this
phase error is averaged and plotted versus the ratio of the section, We assume the conventional methods for comparisons
hidden neuron number to the input terminal number JML /IML . that have been described in Section IV.
The iteration number of ML-CVNN-weight update is 10 for
a single set of inputs. The PPE decreases with the increase
of the ratio, and is minimum at around JML /IML = 1. But, B. Prediction Accuracy
this relationship between IML and JML has less influence
than other parameters because the fluctuation in the range of We evaluate the performance of the prediction methods for
460–490 is relatively small. the measured channel characteristic by assuming the same
OFDM communications with the QPSK modulation as that
has been described in Section IV. We predict the characteristic
V. E XPERIMENTS IN ACTUAL E NVIRONMENT
in a unit TDD frame (5 ms). The window length n is 8
A. Experiment Setup again. Among various cases, we present a set of prediction
In this section, we present the experimental results of result in which the conventional prediction methods fail.
the proposed predictions in actual fading environment. Fig. 15 are the results. The CZT-ML-CVNN predicts the
We conducted experiments for fading channels in actual com- channel characteristic with higher accuracy, whereas the other
munications environment shown in Fig. 12. There are a mobile methods deviate from the actual channel. This fact shows the
user (MU) as a transmitter and a BS as a receiver in the CZT-ML-CVNN-based prediction has higher performance also
experiment site. Table V lists the experiment parameters. The in the actual fading environment. The BER shown in Fig. 16
distance between the MU and the BS is ∼300 m, and the MU also indicates that the CZT-ML-CVNN-based prediction has a
moves in the direction of the arrow shown in Fig. 12 with a higher performance than the conventional methods.

Authorized licensed use limited to: Rohde & Schwarz. Downloaded on February 18,2020 at 13:00:00 UTC from IEEE Xplore. Restrictions apply.
1694 IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, VOL. 25, NO. 9, SEPTEMBER 2014

Fig. 17. Number of communications failure (cumulative phase error

>1500 rad·symbol) in actual fading environment.

Fig. 15. Predicted (a) relative power and (b) phase curves of the channel as
functions of time in the prediction period. VI. C ONCLUSION
We proposed a channel prediction method based on the
CVNNs. The combination of the ML-CVNN and CZT brings
a highly accurate channel prediction even in a low frequency
resolution condition. The channel characteristic has not only
amplitude information, but also phase information. Since the
dominant part of channel changes of respective paths is
rotation in the complex plane, the CVNNs can bring about
both high flexibility and low generalization error in channel
prediction. We deal with respective characteristics as complex
entities. A series of simulations and experiments demon-
strated the communications with the CZT-ML-CVNN-based
prediction could realize better performance than that with the
TD-linear and TD-RTRL-CVNN-based prediction as well as
the CZT-combined linear, Lagrange-based, AR-model-based,
and RTRL-CVNN-based predictions.

ACKNOWLEDGMENT
The authors would like to thank the anonymous reviewers
for their insightful suggestions.
Fig. 16. BER curves obtained for TD-linear prediction, TD-RTRL-CVNN-
based prediction, CZT only, CZT-Linear prediction, CZT-Lagrange-based
prediction, CZT-AR-model-based prediction, CZT-RTRL-CVNN-based pre- R EFERENCES
diction, CZT-ML-CVNN-based prediction, and actual channel characteristics. [1] W. C. Jakes, Microwave Mobile Communications, 2nd ed. New York,
NY, USA: Wiley, 1994.
[2] A. Duel-Hallen, “Fading channel prediction for mobile radio adaptive
transmission systems,” Proc. IEEE, vol. 95, no. 12, pp. 2299–2313,
C. Communications-Failure Rate Dec. 2007.
[3] High Capacity Spatial Division Multiple Access HC-SDMA, ANSI-ATIS
We conducted channel prediction for 70 trials by shifting the Standard 0700004-2005, Sep. 2005.
prediction period with 1-ms step in the rapid fading interval of [4] A. J. Paulraj, D. A. Nabar, and H. Bolcskei, “An overview of MIMO
Fig. 14. The phase error in the predicted phase is accumulated communications—A key to gigabit wireless,” Proc. IEEE, vol. 92, no. 2,
pp. 198–218, Feb. 2004.
in a unit frame. The Hann window length n is 8 again. Fig. 17 [5] F. Maehara, F. Sasamori, and F. Takahata, “Linear predictive maximal
represents the number of prediction with phase error being ratio combining transmitter diversity for OFDM-TDMA/TDD systems,”
>1500 rad·symbol, which means a failure of communica- IEICE Trans. Communicat., vol. E86-B, no. 1, pp. 221–229, 2003.
[6] H. Bui, Y. Ogawa, T. Nishimura, and T. Ohgane, “Performance eval-
tions in the OFDM system. It is found that the number uation of a multi-user MIMO system with prediction of time-varying
of communications failure in the CZT-ML-CVNN is smaller indoor channels,” IEEE Trans. Antennas Propag., vol. 61, no. 1,
than those in other methods. The CZT-ML-CVNN-based pre- pp. 371–379, Jan. 2013.
[7] T. Eyceoz, A. Duel-Hallen, and H. Hallen, “Prediction of fast fading
diction is highly effective also in actual fading-environment parameters by resolving the interference pattern,” in Proc. Asilomar
experiment. Conf. Signals, Syst., Comput., vol. 1. 1997, pp. 167–171.

Authorized licensed use limited to: Rohde & Schwarz. Downloaded on February 18,2020 at 13:00:00 UTC from IEEE Xplore. Restrictions apply.
DING AND HIROSE: FADING CHANNEL PREDICTION BASED ON COMBINATION OF CVNNs AND CZT 1695

[8] T. Eyceoz, A. Duel-Hallen, and H. Hallen, “Deterministic channel mod- [24] A. Hirose and S. Yoshida, “Generalization characteristics of complex-
eling and long range prediction of fast fading mobile radio channels,” valued feedforward neural networks in relation to signal coherence,”
IEEE Trans. Commun. Lett., vol. 2, no. 9, pp. 254–256, Sep. 1998. IEEE Trans. Neural Netw. Learn. Syst., vol. 23, no. 4, pp. 541–551,
[9] A. Arredondo, K. Dandekar, and G. Xu, “Vector channel modeling Apr. 2012.
and prediction for the improvement of downlink received power,” IEEE [25] G. Kechriotis and E. S. Manolakos, “Training fully recurrent neural
Trans. Commun., vol. 50, no. 7, pp. 1121–1129, Jul. 2002. networks with complex weights,” IEEE Trans. Circuits Syst., Analog
[10] M. Sternad and D. Aronsson, “Channel estimation and prediction for Digital Signal Process., vol. 41, no. 3, pp. 235–238, Mar. 1994.
adaptive OFDM downlinks,” in Proc. IEEE Veh. Technol. Conf., vol. 2.
Oct. 2003, pp. 1283–1287.
[11] W. Liu, L.-L. Yang, and L. Hanzo, “Recurrent neural network based Tianben Ding received the B.Eng. degree from
narrowband channel prediction,” in Proc. IEEE Veh. Technol. Conf., Yokohama National University, Kanagawa, Japan, in
vol. 5. May 2006, pp. 2173–2177. 2012. He is currently pursuing the M.S. degree in
[12] C. Potter, G. K. Venayagmoorthy, and K. Kosbar, “RNN based MIMO electrical engineering and information systems from
channel prediction,” Signal Process., vol. 90, pp. 440–450, Feb. 2010. the University of Tokyo, Tokyo, Japan.
[13] J. Hwang and J. Winters, “Sinusoidal modeling and prediction of fast His current research interests include neural net-
fading processes,” in Proc. IEEE Globecom, Nov. 1998, pp. 892–896. works and wireless electronics.
[14] J. Andersen, J. Jensen, S. Jensen, and F. Frederiksen, “Prediction of
future fading based on past measurements,” in Proc. IEEE 50th Veh.
Technol. Conf., 1999, pp. 151–155.
[15] L. R. Rabiner, R. W. Schafer, and C. Rader, “The chirp
z-transform algorithm,” IEEE Trans. Audio Electroacoust., vol. 17, no. 2, Akira Hirose (F’13) received the Ph.D. degree in
pp. 86–92, Jun. 1969. electronic engineering from the University of Tokyo,
[16] S. Tan and A. Hirose, “Low-calculation-cost fading channel prediction Tokyo, Japan, in 1991.
using chirp z-transform,” Electron. Lett., vol. 45, no. 8, pp. 418–420, He was a Research Associate with the Research
Apr. 2009. Center for Advanced Science and Technology
[17] S. Ozawa, S. Tan, and A. Hirose, “Channel prediction experiment based (RCAST), University of Tokyo, in 1987. In 1991,
on linear prediction in frequency domain,” in Proc. Asia-Pacific Microw. he was an Instructor with RCAST. From 1993
Conf., Dec. 2010, pp. 1280–1283. to 1995, he was with the University of Tokyo,
[18] S. Ozawa, S. Tan, and A. Hirose, “Errors in channel prediction based on Institute for Neuroinformatics, University of Bonn,
linear prediction in the frequency domain: A combination of frequency- Bonn, Germany. He is currently a Professor with
domain and time-domain techniques,” URSI Radio Sci. Bull., vol. 337, the Department of Electrical Engineering and Infor-
pp. 25–29, Apr. 2011. mation Systems, University of Tokyo. His current research interests include
[19] H. Matsui and A. Hirose, “Nonlinear prediction of frequency-domain neural networks and wireless electronics.
channel parameters for channel prediction in fading and fast doppler- Dr. Hirose is a Senior Member of the Institute of Electronics, Information,
shift change environment,” in Proc. Int. Symp. Antennas Propag., 2012, and Communication Engineers (IEICE) and a member of the Japanese Neural
pp. 1132–1135. Network Society (JNNS). He served as the Editor-in-Chief of the IEICE
[20] A. Hirose, Complex-Valued Neural Networks, 2nd ed. New York, NY, Transactions on Electronics, an Associate Editor of journals such as the
USA: Springer-Verlag, 2012. IEEE T RANSACTIONS ON N EURAL N ETWORKS and the IEEE G EOSCIENCE
[21] S. Chen, S. McLaughlin, and B. Mulgrew, “Complex-valued radial basis AND R EMOTE S ENSING N EWSLETTER, the Chair of the Neurocomputing
function network, Part I: Network architecture and learning algorithms,” Technical Group, IEICE, and the General Chair of the 2013 Asia-Pacific
Signal Process., vol. 35, no. 1, pp. 19–31, 1994. Conference on Synthetic Aperture Radar, Tsukuba, Japan. He serves as a
[22] S. Chen, S. McLaughlin, and B. Mulgrew, “Complex-valued radial basis member of the IEEE Computational Intelligence Society Neural Networks
function network, Part II: Application to digital communications channel Technical Committee, the Founding Chair of the Complex-Valued Neural
equalisation,” Signal Process., vol. 36, no. 2, pp. 175–188, 1994. Network Task Force in NNTC, the Governing Board Member of the Asia-
[23] A. Hirose, “Applications of complex-valued neural networks to coherent Pacific Neural Network Assembly, the Co-Editor-in-Chief of the Journal of
optical computing using phase-sensitive detection scheme,” Inf. Sci. IEICE, the Vice President of the IEICE Electronics Society, and the President
Appl., vol. 2, no. 2, pp. 103–117, 1994. of the JNNS.

Authorized licensed use limited to: Rohde & Schwarz. Downloaded on February 18,2020 at 13:00:00 UTC from IEEE Xplore. Restrictions apply.