Classify gaze events with a lightweight unsupervised HMM — classify_gazepoint_events

Estimates gaze velocity, initializes hidden states by k-means, estimates a Gaussian-emission HMM, and decodes the most likely sequence with Viterbi. This is a lightweight package-internal HMM classifier and not a replacement for validated laboratory event-detection software.

Usage

classify_gazepoint_events_hmm(
  data,
  x,
  y,
  time,
  subject = NULL,
  n_states = 3,
  state_labels = NULL
)

Arguments

data: A data frame.
x: X-coordinate column.
y: Y-coordinate column.
time: Time column.
subject: Optional subject column for within-subject sequences.
n_states: Number of hidden states.
state_labels: Optional labels for states. If NULL, states are ordered by increasing mean velocity.

Value

The input data with velocity, hmm_state, and hmm_event columns.