LPTime (version 1.0-2)

LPiTrack: Algorithm for eye-movement signal processing

Description

Implements a generic nonparametric statistical algorithm to analyze eye-movement trajectory data.

Usage

LPiTrack(xy_mat, m = c(3, 5, 15), p = 10)

Arguments

xy_mat

A matrix with first column as $x$-coordinates and second column as $y$-coordinates of trajectory data.

m

A vector of three items. m[1]: The number of orthonormal LP-transformations to implement LPTime function. m[2]: The order of LP comoment matrix, excluding zero-order co-moments. m[3]: Number of LP moments.

p

The lag-order for vector autoregressive model to be fitted to the data to extract temporal features in the data.

Value

A vector representation of LP features for the trajectory data, which can be used as covariates (signatures) for subsequent prediction modelling.

Details

This function simultaneously extracts all Temporal-Spatial-Static features from the trajectory data integrating LPTime, LP.comoment and LP.moment functions. LPTime fits VAR model on the LP transformed series to capture the joint (horizontal and vertical) dynamics of the eye-movement pattern. LP.moment is applied on the series $r(t)$ (where we define $r^2(t) \,=\, X^2(t) + Y^2(t)$), $X(t), Y(t),$ and their first and second order differences to capture the static pattern. LP.comoment is applied on the following three series: $ (r(t), \Delta r(t))$, $(X(t), Y(t))$ and $(\Delta X(t),\Delta Y(t))$ to extract nonparametric copula-based spatial fixation patterns.

References

Mukhopadhyay, S. and Nandi, S. (2015). LPiTrack: Eye Movement Pattern Recognition Algorithm and Application to Biometric Identification.

See Also

LPTime, LP.moment, LP.comoment

Examples

library(LPTime)
data(EyeTrack.sample)
head(LPiTrack(as.matrix(EyeTrack.sample), m = c(4,5, 15), p=3))