3.2.1. Heuristic Methods

Heuristic methods build a series of rules that leverage the cyclic patterns in the time domain to perform walk detection and step counting. The representative algorithms include the multiple threshold method (MT) and the finite state machine (FSM) method. The multiple threshold method, which was proposed by Kim et al. [44], makes use of the cyclic peaks, valleys and thresholds to count steps. The finite state machine (FSM) by Alzantot et al. [35] sets some thresholds in the magnitude to drive an FSM to count steps.

In addition, Randell et al. [45], Bylemans et al. [46] and Ailisto et al. [47] proposed algorithms to detect the step event by finding the consecutive local maxima and minima of the low-pass version of the sensor signal. Beauregard et al. [48] found the positive-going zero-crossing event that indicates the boundaries of each step cycle to count steps. Ying et al. [27] detected the negative peaks that were caused by the heel-strike event to count steps. The correspondence between a peak value and a step was shown in the study of Goyal et al. [49], which finds the peak within one zero-crossing interval when the sensor is placed at the pelvis.
