rocinterp

ROC curve interpolation

Since R2021a

Syntax

ipd = rocinterp(snr,pd,snrq,'snr-pd')

isnr = rocinterp(pd,snr,pdq,'pd-snr')

ipd = rocinterp(pfa,pd,pfaq,'pfa-pd')

ipfa = rocinterp(pd,pfa,pdq,'pd-pfa')

Description

ipd = rocinterp(snr,pd,snrq,'snr-pd') returns the probability of detection (P_d) computed by interpolating a P_d vs. signal-to-noise ratio (SNR) receiver operating characteristic (ROC) curve. If pd is a matrix, the function interpolates each column independently. In this and the next syntax, rocinterp performs linear interpolation after transforming the P_d-axis of the ROC curve using the normal probability scale.

isnr = rocinterp(pd,snr,pdq,'pd-snr') returns the SNR computed by interpolating a P_d vs. SNR ROC curve. If snr is a matrix, the function interpolates each column independently.

ipd = rocinterp(pfa,pd,pfaq,'pfa-pd') returns the P_d computed by interpolating a P_d vs. probability of false alarm (P_fa) ROC curve. If pd is a matrix, the function interpolates each column independently. In this and the next syntax, rocinterp performs linear interpolation after transforming both axes of the ROC curve using a logarithmic scale.

ipfa = rocinterp(pd,pfa,pdq,'pd-pfa') returns the P_fa computed by interpolating a P_d vs. P_fa ROC curve. If pfa is a matrix, the function interpolates each column independently.

Examples

Interpolate Probability of Detection vs. SNR ROC Curve

Compute the probability of detection (${\mathit{P}}_{\mathrm{d}}$) for a Swerling 1 case target given a set of signal-to-noise ratio (SNR) and probability of false alarm values. Express the SNR values in decibels.

SNR = [13.5 14.5];
pfa = [1e-9 1e-6 1e-3];

Compute the ${\mathit{P}}_{\mathrm{d}}$ vs. SNR ROC curves and interpolate them at the SNR values of interest.

[pd,snr] = rocpfa(pfa,'SignalType','Swerling1');

ipd = rocinterp(snr,pd,SNR,'snr-pd');

Plot the ROC curves and overlay the interpolated values.

rocpfa(pfa,'SignalType','Swerling1')
hold on
q = plot(SNR,ipd,'*');
hold off
legend(q,append("P_{fa} = ",string(pfa),", int."),'Location','northwest')

Input Arguments

snr — Signal-to-noise ratio

vector | matrix

Signal-to-noise ratio in decibels (dB), specified as a vector or matrix. If snr is a vector, its values must be unique. If snr is a matrix, then each of its columns must contain unique values.

Data Types: double

snrq — Signal-to-noise ratio query points

vector

Signal-to-noise ratio query points, specified as a vector. All values of snrq must be expressed in dB.

Data Types: double

pd — Probability of detection

vector | matrix

Probability of detection, specified as a vector or matrix. All values of pd must be between 0 and 1. If pd is a vector, its values must be unique. If pd is a matrix, then each of its columns must contain unique values.

Data Types: double

pdq — Probability of detection query points

vector

Probability of detection query points, specified as a vector. All values of pdq must be between 0 and 1.

Data Types: double

pfa — Probability of false alarm

vector | matrix

Probability of false alarm, specified as a vector or matrix. All values of pfa must be between 0 and 1. If pfa is a vector, its values must be unique. If pfa is a matrix, then each of its columns must contain unique values.

Data Types: double

pfaq — Probability of false alarm query points

vector

Probability of false alarm query points, specified as a vector. All values of pfaq must be between 0 and 1.

Data Types: double

Output Arguments

ipd — Interpolated probability of detection

vector | matrix

Interpolated probability of detection, returned as a vector or matrix.

isnr — Interpolated signal-to-noise ratio

vector | matrix

Interpolated signal-to-noise ratio, returned as a vector or matrix.

ipfa — Interpolated probability of false alarm

vector | matrix

Interpolated probability of false alarm, returned as a vector or matrix.

Extended Capabilities

C/C++ Code Generation
Generate C and C++ code using MATLAB® Coder™.

Version History

Introduced in R2021a