Keys for Action: An Efficient Keyframe-Based Approach for 3D Action Recognition Using a Deep Neural Network

. 2020 Apr 15;20(8):2226. doi: 10.3390/s20082226

Algorithm 1 Keyframe Extraction Algorithm

1:
procedureKeyframesExtraction( $M$ )
2:
inputs: $M = {X_{1}, X_{2}, X_{3}, \dots, X_{n}}$ , given input sequence of frames
3:
  persistent: n, total number of frames

  k, total number of nearest neighbors

   $N = {{\hat{X}}_{1}, {\hat{X}}_{2}, {\hat{X}}_{3}, \dots, {\hat{X}}_{k}}$ , retrieved nearest neighbors

   $D = {d_{1}, d_{2}, d_{3}, \dots, d_{k}}$ , distances computed for each nearest neighbor

   $Φ = {ϕ_{1}, ϕ_{2}, ϕ_{3}, \dots, ϕ_{c}}$ , candidate keyframes

   $Ψ = {ψ_{1}, ψ_{2}, ψ_{3}, \dots, ψ_{s}}$ , final selected keyframes

  t, threshold value
4:
$k \leftarrow \frac{n}{2}$
5:
$s \leftarrow 1$
6:
while $(l e n g t h (M) \neq 0)$ do
7:
$X_{1} \leftarrow e x t r a c t F r a m e (M)$
8:
$[N, D] \leftarrow f i n d N e a r e s t N e i g h b o u r s (M, X_{1}, k)$
9:
$i \leftarrow 1$
10:
while $(i < k)$ do
11:
if $(d_{i} < t)$ then
12:
$ϕ_{i} \leftarrow {\hat{X}}_{i}$
13:
$i \leftarrow i + 1$
14:
end if
15:
end while
16:
$Φ \leftarrow s o r t (Φ)$
17:
$ψ_{s} \leftarrow m e d i a n (Φ)$
18:
$M \leftarrow M - Φ$
19:
$s \leftarrow s + 1$
20:
end while
21:
end procedure