Fig. 5: Radar classification experiment using the PEIC.
From: Photonic edge intelligence chip for multi-modal sensing, inference and learning
a Schematic of the experimental setup. It illustrates the radar detection system, where microwave signals are transmitted and reflected by various objects (square, cross, circle, plus). The reflected signals are received and converted into optical spectra via an electro-optic phase modulator. The multi-wavelength laser is used to generate discrete optical lines across a broad spectral range, enabling precise sampling of the radar-modulated optical signal. The PEIC processes these signals for radar classification. b (i) Different objects (square, cross, circle, plus) used to demonstrate radar detection. b (ii) Time-domain waveforms at the receiver, computed via the geometric optics (GO) method, showing distinct signal patterns for each object. b (iii) Experimentally measured spectral responses of the radar-modulated optical signal, revealing each object’s unique spectral signature. c In situ supervised training curves for radar classification, reaching 83.4% accuracy. d Confusion matrices comparing classification results with and without fine-tuning. Accuracy improves from 28.1% to 76.9% through unsupervised fine-tuning.
