Assured Signal
Release: Beta — suitable for trial and evaluation; APIs may change before GA
Bounded-influence signal denoising for defense sensors
Assured Signal is a drop-in replacement for Takens local-projection (LP) denoising in radar, sonar, EW, CBRNE, SIGINT, and autonomous sensor fusion pipelines. It uses adaptive bounded-influence projection with zero manual tuning required.
Validated across a broad benchmark suite against published LP methods — substantially better noise removal at operational SNR levels while preserving high-SNR signals.
What it does
Your sensor pipeline receives a time series — radar returns, sonar pings, SIGINT samples — and the noise is high enough that downstream tracking or classification starts to fail. Standard Takens local-projection (LP) denoising often helps at moderate SNR but collapses at operational noise levels (0 dB and below).
Assured Signal replaces that LP stage. You pass in a noisy window; it returns a cleaned signal. The recommended adaptive method picks processing intensity automatically — no manual tuning of embedding dimension, neighbors, or iteration count. At high SNR it stays conservative so it never damages a clean signal; at low SNR it applies stronger bounded-influence projection.
When to use it
Use Assured Signal anywhere you currently run LP denoising upstream of fusion, ATR, or feature extraction — especially when operators report "works in the lab, fails in the field."
See Examples for runnable code.
Who it is for
- Defense program offices and primes using Takens LP for nonlinear signal denoising
- Radar, sonar, EW, and SIGINT pipelines that fail at operational noise levels
- Embedded and air-gapped deployments requiring on-premise signal preprocessing
What ships
- C library (
libassuredsignal) with public header - Python wheel (
assured-signal) with ctypes binding - Seven denoising methods including the recommended adaptive default