Table 1.

Sensitivity, specificity, and FPR are shown for varying CNRs for the proposed algorithm using $\widehat{p}<0.05$ as the threshold for activation. Specificity is equal to 1 — FPR. Sensitivity is equal to the true positive rate. The offline results for the full 5 min of simulation from the OLS and AR-IRLS models are also presented.

	Kalman model					AR-OLS	AR-IRLS
	1 min	2 min	3 min	4 min	5 min	5 min	5 min
Sensitivity | True positive rate
$\mathrm{SNR}=0.5$	10.8%	28.4%	40.1%	49.3%	56.0%	21.2%	43.5%
$\mathrm{SNR}=1.0$	29.5%	57.7%	71.6%	79.4%	84.0%	40.4%	80.2%
$\mathrm{SNR}=2.0$	57.9%	82.6%	91.7%	95.1%	96.5%	65.5%	94.5%
Specificity
$\mathrm{SNR}=0.5$	97.6%	97.9%	97.4%	97.3%	97.2%	91.3%	98.3%
$\mathrm{SNR}=1.0$	97.6%	97.7%	97.3%	97.3%	97.1%	88.2%	98.3%
$\mathrm{SNR}=2.0$	97.5%	97.7%	97.3%	97.3%	97.2%	88.4%	98.5%
False positive rate
$\mathrm{SNR}=0.5$	2.4%	2.1%	2.6%	2.7%	2.8%	8.7%	1.7%
$\mathrm{SNR}=1.0$	2.5%	2.3%	2.7%	2.7%	2.9%	11.8%	1.7%
$\mathrm{SNR}=2.0$	2.5%	2.3%	2.7%	2.7%	2.8%	11.6%	1.5%