A method is proposed for estimating the number and parameters of sinusoids in a background of AWGN. The method filters the sinusoids by exploiting their deterministic nature. Compared to other methods, it is more efficient and more reliable at low signal-to-noise ratio. The required assumptions are less stringent. The flexibility of the technique allows important modifications such as the possibility of detecting the absence of sinusoids instead of their presence and of extending the technique to the multidimensional case. Simple microprocessor implementation is feasible.
展开▼