Читать книгу Dynamic Spectrum Access Decisions - George F. Elmasry - Страница 34

This spectrum sensing approach has more advantages than the matched filter approach described in the previous section when attempting to detect the presence of a narrow band signal. Autocorrelation estimation can result in a constant false alarm rate (CFAR) by employing techniques that reduce the dependency on noise power. With this sensing technique, the sensor can have a stored representation of the signal as samples S(t) = (s₀, s₁, …, s_{N − 1}) or has a signal generator that can generate a copy of the sensed signal and sample it. The sensed signal S^*(t) is sampled to produce. The result of the correlation process¹⁰ between S(t) and S^*(t) is compared to threshold to decide if the sensed signal matches S(t) or not, as shown in Figure 2.8.

Figure 2.8 Signal detection using autocorrelation.

If the sensor uses a signal generator and a sampler instead of stored samples, as in Figure 2.8, the sampler can be broken down as shown in Figure 2.9 where down conversion and a low pass filter (LPF) is used before time sampling.¹¹ In Figure 2.9, the signal is a complex baseband signal with center frequency f_c. The LPF has a bandwidth of (−f_bw, f_bw) Hz and the sampling rate is.

Figure 2.9 Signal sampling before autocorrelation.

The correlation technique attempts to produce unbiased estimation of the signal that can be expressed in terms of a variable l that is adaptively chosen to reduce noise power dependency. Different techniques can be employed to incorporate l in the correlation function, such as estimating a covariance matrix.

Autocorrelation based spectrum sensing can use a bank of signal generators, samplers, and correlators to detect the presence of multiple signals.

Notice the difference between time‐domain cyclic marks and frequency domain cyclic marks. OFDM signals use a frequency‐domain cyclic prefix, which protects the OFDM signals¹² from inter‐symbol interference. This cyclic prefix can be utilized in a frequency‐domain based correlation technique to affirm the presence of the targeted signal. If the targeted signal uses a preamble of symbols, this preamble can be utilized to affirm the presence of the signal in time‐domain correlation. Both cyclic prefix and time‐domain preamble can help the autocorrelation capable sensor decrease the probability of misdetection and the probability of false alarm when making a decision regarding the presence of the targeted signal.