Intelligent Video Highlights Generation with Front-Camera Emotion Sensing

. 2021 Feb 3;21(4):1035. doi: 10.3390/s21041035

Algorithm 2 Sound events clustering

Require: sub_events D, $Δ t_{t h}$ , $Δ y_{t h}$ , $T_{m a x}$

Ensure: events E

1:
y = amplitude of a sub-event/event
2:
X ← first cluster to be grown, initialized with $d_{0}$
3:
for each sub_event $d_{i}$ of D do
4:
if ( $t_{X} - t_{d_{i}}$ < $Δ t_{t h}$ ) & ( $| y_{X} - y_{d_{i}} |$ < $Δ y_{t h}$ ) then
5:
X ← X + $d_{1}$
6:
y ← new weighted amplitude of X
7:
else
8:
insert $d_{i}$ into E as a new cluster to be grown
9:
end if
10:
end for
11:
for each pair ( $E_{i}$ , $E_{i + 1}$ ) in E do
12:
$E_{l o w}$ ← event of ( $E_{i}$ , $E_{i + 1}$ ) with lowest score
13:
if ( $t_{E_{i}} - t_{E_{i + 1}}$ < $Δ t_{t h}$ ) & ( $| y_{E_{i}} - y_{E_{i + 1}} |$ < $Δ y_{t h}$ $| |$ $T_{E_{l o w}} < T_{m a x}$ ) then
14:
Merge $E_{i}$ with $E_{i + 1}$ , the new y being weighted with respect to events widths
15:
end if
16:
end for