Articulated Human Motion Tracking Using Sequential Immune Genetic Algorithm

Algorithm 3

S-IGA-based motion tracking algorithm.

(1) Initialization: perform human pose estimation on the first frame of the video, output the

converged antibodies {𝑥_1,𝑡(𝐾),𝑥_2,𝑡(𝐾),…,𝑥_𝑁,𝑡(𝐾)}, where 𝑡=0;

(2) for  𝑡=1∶𝑇  do

(3) Next-frame propagation: randomly propagate the antibodies to enhance their diversities

according to the following transition model:

   𝑥_𝑖,𝑡(0)∼𝑁(𝑥_{𝑖,𝑡−1}(𝐾),Σ),  𝑖=1,…,𝑁;

(4) IGA-based pose optimization: using Algorithm  2 to optimize the initial antibodies:

   {𝑥_1,𝑡(0),𝑥_2,𝑡(0),…,𝑥_𝑁,𝑡(0)};

(5) Check the convergence criterion: if satisfied, the converged antibodies are used to initial

the next frame;

(6) The individual with best affinity in population {𝑥_1,𝑡(𝐾),𝑥_2,𝑡(𝐾),…,𝑥_𝑁,𝑡(𝐾)} is used to

approximate the tracking result of time 𝑡;

(7) end for.