Detection: below-noise structured signal (entropy scan).
hypothesisdetection_strategyScientific Instrumentation & Sensor Data
Encode information beneath a channel's noise floor; detect via entropy/compressibility/matched-filter scans of large natural datasets for anomalously low-entropy 'noise'.
Relaxes the assumption a real signal must exceed per-sample SNR (spread-spectrum hides recoverable signals below noise). Our detectors threshold per-event; sub-noise coherent structure needs matched-filter analysis rarely run on natural data. HYPOTHESIS. [DEEP DIVE] Now a runnable, false-positive-controlled prototype: research/detection/below_noise_entropy_scan.py recovers a sub-noise coherent signal (per-sample SNR 0.13) via a trials-corrected periodogram threshold (selftest: FPR 0.045, detect-rate 0.95). Falsification contract: null=white-noise Exp(1) bins; artificial signature=coherent power the null cannot produce at threshold; retire if no bin exceeds threshold after real-null calibration + line-notching. Limit: recovers a signal only if one is actually embedded (the coupling limit).