Algorithm 1 Step Detection Algorithm.

Input: Filtered acceleration norm ${\tilde{y}}_a$, threshold value $B_{th}$, window length $N_s$.     Output: Double stance indices $P_H$ at high peak of acceleration norm. 1:Compute the number of samples N of ${\tilde{y}}_a$ 2:for discrete time k from $k=N_s+1$ to $N-N_s$ do 3: if ${\tilde{y}}_{a,k}>B_{th}$ and ${\tilde{y}}_{a,k}\ge max({\tilde{y}}_{a,k-N_s}:{\tilde{y}}_{a,k-1})$ and ${\tilde{y}}_{a,k}\ge max({\tilde{y}}_{a,k+1}:{\tilde{y}}_{a,k+N_s})$ then 4:  $P_H=[P_H,k];$ 5: end if 6:end for