on/off detection quality measures {biosignalEMG}

Measure the quality of an automatic on/off detection

Description

These quantities measure different aspects of the accuracy in the determination of activity changes of an electric signal.

Usage

ANDP(b, bE)
MNChPD(b, bE)
PCE(b, bE)
PR(b, bE, t)
TD(b, bE, t)

Arguments

b

reference vector (target values) indicating for each position if the datum corresponds to a silence phase (0) or an active phase (1 or a greater integer indicating the level of activation).

bE

detected vector (output of an automatic detector) indicating for each position if the datum corresponds to a silence phase (0) or an active phase (1 or a greater integer indicating the level of activation).

t

tolerance value (see Details).

Details

These quality measures are defined as

ANDP	The absolute difference of the number of detected phases and the actual number of pahses
MNChPD	The mean of the distances of each detected change-point to the nearest actual change-point
PCE	The percentage of incorrectly classified points (silence-activity)
TD	temporal deviation
PR	computes the true positive ratio (TPR) and the false positive ratio (FPR).

TD and PR depends on the value of t, which is a tolerance for the difference between the calculated and exact changepoints.

See Guerrero et.al. (2014) for details on the computation of these measures.

Value

ANDP, MNChPD, PCE, TD: a numeric value. PR: a list of two numeric values (TPR and FPR).

Note

The parameter t should be adjusted in terms of the sampling rate of the EMG.

Author(s)

J.A. Guerrero [e-mail]

References

Guerrero J.A., Macias-Diaz J.E. (2014) A computational method for the detection of activation/deactivation patterns in biological signals with three levels of electric intensity. Math. Biosci. 248, 117–127.

Pistohl T., Schmidt T.S.B., Ball T., Schulze-Bonhage A., Aertsen A., Mehring C. (2013) Grasp detection from human ECoG during natural reach-to-grasp movements. PLoS ONE 8

See Also

onoff_bonato, onoff_singlethres

Examples

# Simulate 10 seconds of an EMG
emgx <- syntheticemg(n.length.out = 10000, on.sd = 1, on.duration.mean = 350, 
    on.duration.sd = 10, off.sd = 0.05, off.duration.mean = 300, off.duration.sd = 20, 
    on.mode.pos = 0.75, shape.factor = 0.5, samplingrate = 1000, units = "mV", 
    data.name = "Synthetic EMG")

# Detect the phases of activation in emgx
b_bonato <- onoff_bonato(emgx, sigma_n = 0.05, m = 10, minL = 30)
b_singlet <- onoff_singlethres(emgx, t = 0.2)

# Compute the quality measures
qm_bonato <- c(ANDP(b_bonato, emgx$on.off), MNChPD(b_bonato, emgx$on.off), 
    PCE(b_bonato, emgx$on.off), PR(b_bonato, emgx$on.off, t = 10), TD(b_bonato, 
        emgx$on.off, t = 10))

qm_singlet <- c(ANDP(b_singlet, emgx$on.off), MNChPD(b_singlet, emgx$on.off), 
    PCE(b_singlet, emgx$on.off), PR(b_singlet, emgx$on.off, t = 10), TD(b_singlet, 
        emgx$on.off, t = 10))

res <- as.matrix(cbind(qm_bonato, qm_singlet))
rownames(res) <- c("ANDP", "MNChPD", "PCE", "TPR", "FPR", "TD")
print(res)