Sequential test and parameter estimation for array processing of seismic data

We present a sequential test and parameter estimation technique for measured seismic data from the GERESS array situated in Germany. A new approximation of the test statistic distribution and the test threshold is proposed. The sequentially rejecting Bonferroni-Holm test guarantees a global test level. This allows one to avoid the computationally expensive bootstrap method and leads to a simpler algorithm. Approximate conditional maximum likelihood estimates in the frequency domain are used to overcome the resolution limits of conventional methods for wideband signal processing and to construct the sequential test. The combination of global optimization via genetic algorithm and a local one using the scoring seems to be a good compromise to handle the problem of cumbersome maximization of the log-likelihood function over the parameters of interest. The algorithm for testing the number of signal phases is applied simultaneously with the estimation of the model parameters