DVR-Based Detection: Better Than Matched Filter

This new detection algorithm uses local projections to significantly improve pulse detection in noisy environments.

✅ 25–28 dB SNR improvement over matched filter
✅ Local noise adaptation (not fooled by overlap)
✅ Mathematically distinct from simple nonlinear filters

Test performed under high noise to show that DVR is structurally different — not just a nonlinear transformation like exp(MF), tanh(MF), or softplus(MF).

2. Overlapping Pulses – Narrowband Case

Matched filter creates a falsely amplified peak. DVR remains true to signal structure and suppresses false energy merging.

DVR Enhances Kalman Filter Accuracy

Dynamic Variational Reconstruction (DVR) can be used as a post-processing step to significantly improve Extended Kalman Filter (EKF) estimates.

In our benchmark, DVR reduced RMSE to 0.049, outperforming classical methods like Savitzky–Golay smoothing (0.059) and spline interpolation (0.29).