extract_saccades {saccadr}

Extract saccades from samples using votes from selected methods.

Description

Extract saccades from samples using votes from selected methods. Each method votes whether a given sample belongs to a saccade. Next, saccades are identified via a majority vote using the vote_threshold parameter, as well as a minimum duration and minimal temporal separation criteria. Please note that units of the gaze samples must be in degrees of visual angle. The units are important as some methods use specific (e.g., physiologically plausible) velocity and acceleration thresholds.

By default, ensemble includes methods proposed by Engbert & Kliegl (2003) ("ek"), Otero-Millan et al. ("om"), and Nyström & Holmqvist (2010) ("nh"), see Implemented Methods vignette. However, it can be extended via custom methods, see Using Custom Methods vignette.

By default, the function returns a table with identified saccades but can return a matrix with methods' votes per sample instead (return_votes = TRUE).

Usage

extract_saccades(
  x,
  y,
  sample_rate,
  trial = NULL,
  methods = list(method_ek, method_om, method_nh),
  velocity_function = saccadr::diff_ek,
  options = NULL,
  binocular = "merge",
  vote_threshold = ifelse(length(methods) == 1, 1, (length(methods) - 1)),
  minimal_duration_ms = 12,
  minimal_separation_ms = 12,
  return_votes = FALSE
)

Arguments

x

Horizontal coordinate, either a vector for monocular data or a two-column matrix for binocular data.

y

Vertical coordinate, either a vector for monocular data or a two-column matrix for binocular data.

sample_rate

Sampling rate in Hz. It is assumed to be common for the entire time series. If the time series contains chunks (trials) that were recorded using different acquisition rate (e.g., SR Research Eyelink allows to set different acquisition rate for each recording / trial), you would need to split the time series and analyze them separately.

trial

Optional vector with trial ID. If omitted, all samples are assumed to belong to a single trial. Velocity, acceleration, and saccades themselves are computed respecting trial borders.

methods

A list with saccade detection methods, can include external functions that implement sample classification (see Using Custom Methods vignette). Package methods include Engbert & Kliegl (2003) (method_ek), Otero-Millan et al. (2014) (method_om), Nyström and Holmqvist (2010) (method_nh). Defaults to the list of all internally implemented methods: list(method_ek, method_om, method_nh).

velocity_function

A handle to a function to compute velocity and acceleration. Defaults to a method suggested by Engbert & Kliegl (2003) diff_ek. The package also implements the method proposed by Nyström and Holmqvist (2010) diff_nh. See vignette "Velocity computation" for details and information on how to implement a custom method.

options

A named list with options for saccade detection (see method_ek, method_om, method_nh) and velocity (diff_ek, diff_nh) computation. See documentation on specific method for details.

binocular

Specifies how a binocular data is treated. Options are "cyclopean" (binocular data is converted to an average cyclopean image before saccades are extracted), "monocular" (saccades are extracted independently for each eye), "merge" (default, sample votes are obtained from both eyes and for all methods and then averaged. This way only binocular saccades, i.e., eye movements with a sufficient temporal overlap between eyes, are detected.).

vote_threshold

Value between 1 and N (where N is number of used methods) defining a vote threshold for a saccade. By default, all but one method threshold = N-1 must agree for a sample to be considered for a saccade. Threshold of 1 is applied if a single method is used.

minimal_duration_ms

Minimal duration of a saccade in milliseconds. Shorter candidate saccades are discarded,

minimal_separation_ms

Minimal time separation between saccades in milliseconds. Saccades that are separated by a shorter interval of "not a saccade" votes, will be merged including that period.

return_votes

Logical. Whether function should return extracted microsaccades (FALSE, default) or votes per sample (TRUE).

Details

Variables that describe saccade

• Trial Trial index.

• Eye "Monocular" for monocular inputs. "Cyclopean" for binocular data that was averaged before applying algorithms. "Binocular" for binocular data with votes averaged after applying algorithms. "Left" or "Right" for binocular data when eyes are processed independently.

• OnsetSample Index of the first sample.

• OffsetSample Index of the last sample.

• Onset Onset time relative to the trial start in milliseconds.

• Offset Offset time relative to the trial start in milliseconds.

• Duration Duration in milliseconds.

• DisplacementX Horizontal displacement measured from the first to the last sample.

• DisplacementY Vertical displacement measured from the first to the last sample.

• Displacement Displacement magnitude measured from the first to the last sample.

• DisplacementPhi Displacement direction measured from the first to the last sample.

• AmplitudeX Horizontal displacement measured from the leftmost to the rightmost sample.

• AmplitudeY Vertical displacement measured from the lowest to the uppermost sample.

• Amplitude Displacement magnitude measured from the most extreme samples.

• AmplitudePhi Displacement direction measured from the most extreme samples.

• VelocityPeak Peak velocity.

• VelocityAvg Average velocity.

• AccelerationPeak Peak acceleration.

• AccelerationAvg Average acceleration.

• AccelerationStart Peak acceleration before peak velocity was reached.

• AccelerationStop Peak acceleration after peak velocity was reached.

Value

A data.frame with saccade properties (see details), if return_votes = FALSE. Alternatively, it returns votes per sample (return_votes = TRUE). For a monocular processing (monocular input, cyclopean or merged binocular data) it is a matrix with nrow(x) rows and length(methods) columns with 0/1 votes for each sample and method. For binocular processing, function returns a two element list with the similar matrices but per eye.

See Also

method_ek, method_om, method_nh, diff_ek, diff_nh

Examples

# Single trial
data(single_trial)
saccades <- extract_saccades(single_trial$x, single_trial$y, 500)

# Multiple trials
data(monocular_ten_trials)
saccades <- extract_saccades(monocular_ten_trials$x,
                             monocular_ten_trials$y, 
                             500,
                             trial = monocular_ten_trials$trial)
 
 # binocular saccades                            
 data("single_trial_binocular")
 saccades_b <- saccadr::extract_saccades(single_trial_binocular[, c('xL', 'xR')],
                                         single_trial_binocular[, c('yL', 'yR')],
                                         sample_rate = 1000)
                                         
 # cyclopean saccades from binocular data
saccades_c <- saccadr::extract_saccades(single_trial_binocular[, c('xL', 'xR')],
                                        single_trial_binocular[, c('yL', 'yR')],
                                        sample_rate = 1000,
                                        binocular = "cyclopean")

 # monocular saccades from binocular data
saccades_m <- saccadr::extract_saccades(single_trial_binocular[, c('xL', 'xR')],
                                       single_trial_binocular[, c('yL', 'yR')],
                                       sample_rate = 1000,
                                       binocular = "monocular")
                             
# Using a single method
saccades <- extract_saccades(single_trial$x, single_trial$y, 500, methods = method_om)

# Using two methods
saccades <- extract_saccades(single_trial$x,
                             single_trial$y,
                             500,
                             methods = list(method_ek, method_om))

#  Alternative velocity computation method
saccades <- extract_saccades(single_trial$x, single_trial$y, 500, velocity_function = diff_nh)

# A strict unanimous decision threshold
saccades <- extract_saccades(single_trial$x, single_trial$y, 500, vote_threshold = 3)

# A slacker criterion that at least one of the three methods must label sample as a saccade
saccades <- extract_saccades(single_trial$x, single_trial$y, 500, vote_threshold = 1)

# Only longish saccades are extracted
saccades <- extract_saccades(single_trial$x, single_trial$y, 500, minimal_duration_ms = 20)