Recognizing facial emotions for educational learning settings

Akputu Oryina Kingsley, Udoinyang G. Inyang, Ortil Msugh, Fiza T. Mughal, Abel Usoro

Abstract


Educational  learning  settings  exploit  cognitive  factors  as  ultimate  feedback  to  enhance  personalization  in  teaching  and  learning.  But  besides  cognition,  the  emotions  of  the  learner  which  reflect  the  affective  learning  dimension  also  play  an  important  role  in  the  learning  process.  The  emotions  can  be  recognized by tracking  explicit behaviors of  the  learner  like  facial  or vocal  expressions.  Despite  reasonable  efforts  to  recognize  emotions,  the  research  community  is  currently  constraints  by  two  issues,  namely : i)  the  lack  of  efficient   feature   descriptors   to   accurately   represent   and   prospectively  recogniz e (detecting)  the  emotions  of  the  learner ; ii)  lack  of  contextual  datasets to benchmark performances of emotion recognizers in the learning - speci fic  scenarios,  resulting  in  poor  generalizations.  This  paper  presents  a  facial emotion recognition technique  (FERT). The FERT is realized through  results  of  preliminary  analysis  across  various  facial  feature  descriptors.  Emotions  are  classified  using  the  m ultiple  kernel  learning  (MKL)  method  which  reportedly  possesses  good  merits.  A  contextually  relevant  simulated  learning emotion  ( SLE ) dataset is introduced to validate the FERT scheme.  Recognition performance of  the  FERT scheme generalizes to 90.3% on the  SLE  dataset.  On  more  popular  but  noncontextually datasets,  the  scheme  achi e ved 90.0% and 82.8% respectively  extended  Cohn Kanade (CK+) and  acted  facial  expressions  in  the  wild ( AFEW ) datasets.  A  test  for  the  null  hypothesis   that   there   is  no   significant   difference   in   the   performances  accuracies  of  the  descriptors  rather  proved  otherwise  ( x2 = 14 . 619 , df = 5 , p = 0 . 01212 ) for a model considered at  a  95% confidence interval.

Keywords


Con textual dataset; Educational learning; Emotions recognition; Learning emotions; Personalization

Full Text:

PDF


DOI: http://doi.org/10.11591/ijra.v11i1.pp21-32

Refbacks

  • There are currently no refbacks.


Creative Commons License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

IJRA Visitor Statistics