Maximum likelihood estimation of sinusoidal parameters using a global optimization algorithm

We address the problem of determining maximum likelihood (ML) estimates of sinusoidal parameters. Our approach is to use an interval method, a global optimization algorithm, to determine the maximum of the likelihood function. In contrast, existing ML methods use gradient-based optimization algorithm which are known to have problems with local minimum. The interval method algorithm is based on interval arithmetic, which determines a range of values for the unknown parameters instead of a single valve. This property makes it robust to the noise in the data. We present preliminary simulations to demonstrate the performance of the method.