Institusion
Institut Teknologi Perusahaan Listrik Negara
Author
SUNDARI, YOLANDA
Yosrita, Efy
Riki Ruli Affandi, Riki Ruli
Subject
Teknik Informatika
Datestamp
2023-06-06 02:20:43
Abstract :
The purpose of research on the application of algorithms Long Short Term Memory for
the development of the Speech Emotion Recognition in Children aged 5-10 years, namely
to produce a model and predictive value because children aged 5 to 10 years recognize
emotions from speech at a poor level so that the majority fail to recognize emotions from
correct speech. The majority of children who have difficulty managing their emotions
often cry, whine, scream, slam or break things. Behavior in these children is a form of
negative behavior and excessive emotion or is called temper tantrum. From these
problems, this study applies a classification method using an algorithm Long Short Term
Memory with the modeling process done over the last 5 months to find the best
architecture. From architectural experiments LSTM, an optimal model is produced with
an accuracy training value of 92.43% and validation accuracy 93.63%. From the tests
that have been carried out by testing confusion matrix, for emotion surprised obtained
value presentation with accuracy 97%, emotion calm with accuracy 97%, emotion happy
with accuracy 98%, emotion sad with accuracy 98%, emotion angry with accuracy 99%,
emotion fearful with accuracy 98%, emotion disgust with accuracy 99%.