14.1 INTRODUCTION

In previous chapters, most of the discussion on spectrum estimation has concentrated on fast Fourier transform (FFT) because it has been used in digital receiver designs. In this chapter, some other spectrum estimation approaches will be introduced that will be referred to as high-resolution spectrum estimations. Their major advantage is that they can provide higher frequency resolution than FFT, especially on simultaneous signals.

If there are two signals with frequencies very close, an FFT operation may generate one peak containing both signals. High-resolution spectrum estimation may separate the two signals by generating two sharp peaks. The major drawback of applying high-resolution spectrum estimation to digital microwave receivers is the complexity of the operation. Because of the large number of operations required to estimate the frequencies, they might not be implemented for real-time application in the near future. However, they might be implemented for special applications in the near future. For example, if the peak of an FFT output appears to contain more than one signal, high-resolution spectrum estimation may be used to find the frequencies. Thus, the operation may not be required on all the input data.

Many different high-resolution approaches can be used to estimate frequencies from digitized input data. In this chapter, seven high-resolution methods will be discussed, and they are as follows:

1. Linear predication (or autoregressive (AR)) method;

2. Prony's method;

3. The least squares Prony's method;

4. The multiple signal classification (MUSIC) method;

5. The estimation of signal parameters via rotational invariance techniques (ESPRIT) method;

6. The minimum norm...