Performance of Prefiltered Model-Based Frequency Estimators

In this work, the performance improvement due to prefiltering of inputs in model-based frequency estimators is investigated based on simulation experiments. Initial estimates on the tone frequency locations, which are obtained via DFT peak picking type preanalysis, are used to design a prefilter to remove noise and interference. The simulations indicate that prefiltering can improve the accuracy of Pisarenko and AR frequency estimators and MUSIC and KT frequency estimators with low subspace order significantly. The SNR thresholds of model-based frequency estimators are lowered by prefiltering. Additionally, interesting trade-offs between prefiltering gain and the gain due to subspace noise filtering have been investigated.

Performance of Prefiltered Model-Based Frequency Estimators

In this work, the performance improvement due to prefiltering of inputs in model-based frequency estimators is investigated based on simulation experiments. Initial estimates on the tone frequency locations, which are obtained via DFT peak picking type preanalysis, are used to design a prefilter to remove noise and interference. The simulations indicate that prefiltering can improve the accuracy of Pisarenko and AR frequency estimators and MUSIC and KT frequency estimators with low subspace order significantly. The SNR thresholds of model-based frequency estimators are lowered by prefiltering. Additionally, interesting trade-offs between prefiltering gain and the gain due to subspace noise filtering have been investigated.