Encoding the World into the Brain Frequencies: A Neural Vision Framework
A visual neural-input framework for representing visual information through structured neural-frequency encoding.
Abstract
Non-invasive brain computer interfaces have advanced in decoding motor and perceptual activity, but sending immersive sensory and visual information back into the brain in a form it naturally understands remains unsolved. This study explores an AI driven framework for translating visual scenes into neural frequency templates which aligns with the oscillatory dynamics of the human visual system to deliver them directly into the visual cortex. The method begins by processing visual inputs through spatial and temporal frequency decomposition using Fourier and wavelet transforms to extract input across foveated and peripheral fields. These inputs are mapped to oscillatory patterns using AI model trained on electrophysiological data. Resulting templates encode information through cross frequency coupling and phase amplitude modulation which enables targeted entrainment of neural circuits. Output signals are delivered to brain’s visual cortex through immersive stimulation modalities such as tACS, TMS, and tFUS synchronized to user’s endogenous rhythms. Visual evoked potentials and EEG based attention signals provide feedback to the AI model to adapt stimulation in real time which forms a bidirectional communication. Early theoretical models suggests that structured oscillatory stimulation can encode visual information in a brain compatible format which enables visual stimulation without any visual input from eyes. This study provides a foundation for immersive BCI technologies which sends visual information directly to brain, enabling breakthroughs in virtual reality, neurorehabilitation, healthcare, and human computer interactions.
Overview
This research focuses on visual neural input: how visual information might be represented through structured neural-frequency patterns within a noninvasive BCI framework.
Visual encoding
Visual scenes are decomposed into spatial and temporal frequency information across foveated and peripheral fields. Fourier and wavelet features are then mapped to neural-frequency templates using models trained on electrophysiological data, with cross-frequency coupling and phase-amplitude modulation carrying the encoded structure.
Adaptive feedback
Visual evoked potentials and EEG attention signals provide feedback to the mapping system. That response can be used to adapt stimulation timing and patterning to the user’s endogenous rhythms, keeping this study focused specifically on visual information rather than the broader sensory-encoding framework.