Coherence-Geometric Encoder/Decoder: A Multi-Phase Framework for Analog Storage and Transmission

This paper develops a coherence-geometric encoder/decoder architecture for continuous information storage, transmission, perturbation recovery, and reconstruction. Source-dependent representations are mapped into structured multi-phase latent states (μ-space) governed by shared-amplitude and coherence constraints. Channel or storage perturbations act directly on this information-bearing latent representation. Recovery is then performed within the latent geometry through coherence-directed relaxation before the recovered state is decoded and reconstructed in the source domain. The central distinction of the architecture is between latent-state recovery and source reconstruction. The transmitted or stored latent is not treated merely as a passive codeword: its geometry and coherence dynamics participate directly in restoring a perturbed representation prior to reconstruction. The paper demonstrates the framework using spectral audio and image implementations based on paired constant-modulus μ-states. The accompanying computational materials include Jupyter notebooks, source test data, representative reconstructed audio and image outputs, and stored numerical results used to inspect and reproduce the reported experiments. Redundant latent observations are also evaluated by averaging independently perturbed μ-state representations prior to coherence relaxation. The resulting framework provides a practical basis for investigating coherence-geometric approaches to analog and hybrid information transmission, storage, reconstruction, sensing, and continuous information processing. Internal reference: CGI-RSR-000037.