In-fiber high-speed recognition of incoherent-light broadband energy spectrum patterns.

A fiber-optic system is proposed and experimentally demonstrated for real-time, on-the-fly identification of an incoherent-light energy spectrum pattern based on dispersion-induced time-spectrum convolution. In the proposed system, the incoming frequency-spectrum patterns to be identified are modulated by a time-mapped version of the target intensity profile. Following propagation through a suitable fiber-optic dispersive medium, the measured output temporal waveform provides a correlation of the incoming spectra with the programmed target pattern. This enables direct, real-time detection of the matching energy spectra, without any further numerical post-processing. We experimentally demonstrate successful recognition of a target infrared spectral pattern, extending over a bandwidth of 1.5 THz with a resolution of ∼12  GHz, with sub-megahertz update rates. A path for further performance improvements is also suggested.