Algorithm 1: Non-parametric Bayesian clustering algorithm for human motion recognition using collapsed Gibbs sampler
	Input: O⃗(o⃗i, D), Sweeps, Γ(a, b), H⃗
	Output: K, ci
1	begin
2	C⃗←c1, c2, …,cḰ
3	K+←0;
4	foreach s in Sweeps do
5	Cs ← Cs−1
6	foreach s in Sweeps do
7	$m_{-i}\leftarrow \sum _{j=1}^{Ḱ}(c_j==c_i)-1;$
8	if m−i == 0 then
9	cj←cj-1; ∀j ≻ i
10	K+ ← K+ -1
11	end
12	dim ← length(μD)
13	foreach i in K do
14	covariances(i,dim)← iWishart(λ, v)
15	means← MVrnd(μ, covariances(i,dim))
16	$m_k={\sum }_{i=1}^R\mathbb{\text{I}}\left(c_i=k\right)$
17	ci ← MVrnd(means,covariances)
18	end
19	/* Estimate the prior P(o⃗i|O⃗ H⃗) using Equation (31) */
20	if ci ⋏ K+ then
21	K+ ← K++1
22	end
23	/* Estimate the prior P(ci = k|C⃗−i, α) using Equations (23) and (24) */
24	end
25	end
26	/* Collapsed Gibbs Sampling */
27	/* Estimate the posterior P(ci = k|O⃗, C⃗−i, α; H⃗) using Equation (37) */
28
29	end