Abstract
This study proposes Fuzzy-MCUCoder, an adaptive compression method to ensure stable, lifecritical
image transmission in bandwidth- and power-constrained Healthcare IoT environments.
Conventional ultra-lightweight codecs like MCUCoder lack robustness against dynamic network conditions
and image complexity. Fuzzy-MCUCoder extends this model by integrating a fuzzy inference system to
dynamically select the number of latent channels based on communication throughput and image complexity.
This approach resolves the adaptive rigidity of prior work, improving reconstruction quality without
increasing transmission time. Experiments on ImageNet demonstrate that Fuzzy-MCUCoder achieves
significantly more efficient and reliable image transmission with stable visual quality compared to existing
methods. This is particularly critical in remote healthcare scenarios, where missing subtle visual cues such
as posture instability or respiratory anomalies may directly affect clinical decision-making. Consequently, the
proposed method dramatically enhancing reliability in resource-constrained medical IoT infrastructure.
