Cyclic Autocorrelation based Blind OFDM Detection and ...

B. OFDM Signal Identification
As explained before, different OFDM signals can be
classified by the parameters such as the length of guard
interval, subcarrier spacing and so on. In the first step of the
proposed scheme, subcarrier spacing has been calculated after
the OFDM signal is detected. Therefore, another parameter,
length of guard interval is the target of identification in second
step.
Since the duration of useful symbol has been determined in
the first step, we can observe the CAF for =Tu, as shown in
Fig. 1 and Fig. 3. By inserting the =Tu to (5), the test statistic
could be expressed as a function of n, as described in the
following:
2
( )
Z = R T
αn
x u
2
( π N∆fTu) ( π∆
)
N −1
∞
A sin
i2π∆f Tu
2
−i2π( αn−
f ) Tu
= e ⋅ e G( f ) G( α n − f ) df
T sin fT
s u
−∞
The only term determining Z2 is the integration term of
pulse signal. In order to observe the more detail pattern of
CAF, we employ the step size of each as the fractional value
of subcarrier spacing. The smaller step size of each is the
more detailed pattern we can observe. In our simulation, step
size identical to 1/20 of subcarrier spacing is utilized, while the
peaks of CAF are shown in Fig. 3. Therefore, our purpose is to
determine the distance of two maximum peaks in domain,
which is equal to the reciprocal of Ts. We proposed to employ
the cycle frequency searching scheme to determine the two
maximum peaks in the CAF pattern. After calculating the Ts,
we can easily obtain the length of guard interval as Ts-Tu.
IV. SIMULATIONS AND DISCUSSIONS
Simulations are carried out to evaluate the performance of
the proposed blind detection and identification method. OFDM
signal used in the simulations belongs to four modes with
different subcarrier spacing and guard interval. The subcarrier
spacing is expressed in terms of the number of subcarriers in
the same band, such as 1k and 2k. The length of guard interval
is select from 1/4 and 1/8 of useful symbol duration. The
detection and identification performances are evaluated among
these OFDM signals under AWGN.
A. Signal detection performance
Since the first step of signal detection determines the
success of the whole detection and identification process, the
reliability of detection is the main concern when designing the
detection criteria. Peak detection by checking the symmetric
peaks is a good method with low false alarm probability.
Detection performance using the peak detection is evaluated in
terms of the detection and false alarm probabilities under
different SNRs. The results are shown in Fig. 4. It is clear that
larger amount of subcarrier results in better detection
performance. Therefore, 2k subcarrier with 1/4 length of guard
interval achieves the best performance. However, the detection
time should be longer. For the target detection probability of
90%, even the OFDM with shortest symbol duration is able to
be detected below -1dB. This detection performance could be
improved by increasing the observation time, since its symbol
2
(7)
978-1-4244-2108-4/08/$25.00 © 2008 IEEE
duration is shorter than that of other cases. It is also notable
that the false alarm probabilities which are indicated by the
blue curve are almost zeros for all the OFDM signal detection.
It proves that the method by detecting the symmetric peaks is
reliable for the first step.
Figure 5. Simulation results of OFDM signal detection
B. Signal Identification Performance
If the OFDM signal is detected in the first step,
identification is carrier out to determine the symbol duration
using the method described in the previous section. The
performance is evaluated through the probability of successful
identification. Successful identification means the ratio of
guard interval to useful symbol belongs to a certain range of
possible ratio. An example is shown in Fig. 5 with 2 different
OFDM signals. Both of them have 2k subcarriers but different
length of guard interval. In order to guarantee the fairness of
detection with different guard interval, a threshold is select by
6/32 to separate OFDM signals with 1/4 guard interval from
that with 1/8. Meanwhile, a threshold is select by 3/32 to
separate OFDM signals with 1/8 guard interval from that with
1/16. Therefore, the degradation of identification of 1/8 is
mainly due to the shorter symbol duration as well as the
narrower threshold of identification compared to that of 1/4.
Furthermore, this performance could also be improved by
increasing the observation duration.
V. CONCLUSION
Although there are some existing detection methods
proposed for OFDM signal, they require the knowledge either
the number of subcarrier or the spacing between consecutive
subcarriers as a priori. In this paper, we focus on the signal
detection and identification scheme for OFDM system with
unknown parameters. The key parameters to discriminate
different OFDM signals are the subcarrier spacing and length
of guard interval In order to classify different OFDM signals,
a time domain cyclostationarity based approach is proposed
which consists of two steps: first, detect the OFDM signal
from random noise simply by recognizing the symmetric
peaks in the autocorrelations and calculate the subcarrier
spacing; second, by searching the cycle frequencies, the