A Computational WDM Receiver for Wavelength-Selectable Multichannel Visible Light Communication

Compact wavelength-division multiplexing (WDM) receivers have become indispensable in high-capacity optical communication systems due to their compatibility with microelectronic platforms. However, achieving both high-density information transmission and precise optical signal decoding often requires complex fabrication. Moreover, in conventional direct-detection architectures, resolvable channels are limited by the number and wavelength configuration of demultiplexing channels, reducing system flexibility and scalability. Here, we present a computational WDM receiver for visible light communication, incorporating eight Ag/MgF 2 /Pt broadband filter cavities. The streamlined filter design simplifies fabrication and, through its wideband transmission characteristics, efficiently captures multichannel optical signals for spectral data compression. A trained fully connected neural network decodes the intensity and wavelength of arbitrary 1-, 3-, and 16-channel visible spectral signals from the compressed data, achieving an average equivalent crosstalk of −12.6 dB. We further demonstrate a representative application involving the parallel transmission and decoding of string data using three-channel optical signals. This work offers a low-complexity, scalable framework for multichannel visible light communication, advancing compact, high-performance optical systems.