Semiempirical optimization of frequency-diverse holograms for localization

We present a new design technique for frequency-diverse holograms used in millimeter- and submillimeter-wave imaging. A neural-network backed imaging method utilizes quasirandom hologram structure, which disperses the field in the region of interest across the WR-3.4 (220-330 GHz) band. The global optimum for the hologram structure is generally not known. A new, semiempirical optimization method utilizes the experimental data used in localization tasks. The data is a collection of the of the local field at the object plane measured with a probe (S21), and the corresponding localization performance derived from the reflection spectra of a corner cube (S11). The frequency diversity of the local field at the object is correlated with the localization error. The proposed design technique uses the obtained empirical correlation data to inform a new hologram design.