Neural State Inference for Silent Communication

THREE LAYERS OF CAPABILITY

The NEXUS-1 platform detects cognitive and emotional states at standoff distance. The next phase extends this to silent communication.

Subvocalisation Detection

When a person thinks in words, the vocal cords produce micro-activations detectable via surface EMG. MIT Media Lab demonstrated 92% accuracy for a limited vocabulary in 2018. Brainwave Systems is extending to standoff detection via laser vibrometry at 10–50 metres.

Neural Speech Decoding

Invasive BCIs have demonstrated real-time speech decoding at near-conversational rates. The non-invasive equivalent remains below useful accuracy. Current Brainwave approach: high-density EEG (256 channels) combined with fNIRS and transformer-based decoding. Current accuracy: 60–70% for constrained vocabulary.

Direct Neural-to-Text

The Halo Flight Interface pipeline, retrained on linguistic data, can extract pre-motor speech plans. This is a 5–10 year research horizon, not a near-term product.

Integration with the Oracle

Silent communication feeds directly into the Oracle assessment system. An operator wearing the Halo could issue commands, receive intelligence, and coordinate — all without speaking and without any externally detectable communication.