Theoretical analysis of small sample size behaviour of eigenvector projection technique applied to STAP

We investigate finite sample size performance of the eigenvector projection method when applied to space-time adaptive processing (STAP). A theoretical analysis of the expectation of the signal to interference plus noise ratio (SINR) for the eigenvector projection technique is presented. This gives insight into the the problem of determining the optimum choice of the projected clutter subspace. An estimator of the sample-size dependent optimum subspace dimension, which can be significantly smaller than the clutter rank, is also presented. This result, combined with near-optimal eigenvector-free projection techniques with minimal sample support, helps in reducing the computational burden significantly.