Classification of step motion in time-of-flight imaging using the stochastic oscillator and autocorrelation is proposed. Machine learning algorithms correctly identify the step location in 65–75% of trials, with apparent good noise robustness.