Paper Title
Speech Emotional Recognition using Pre-Trained Deep Learning Models
Abstract
To understand human emotion through speech, it is often underestimated that humans have the capability to
understand a basic set of emotions just from how they are spoken or in what tone it is spoken. While on the other hand when
it comes to modern day machines (i.e. computers) it is a task that is stilla very hard process simply because of the many
variables which differ in the sound wave as people talk to each other. A great part in understanding a speech is to understand
the emotion behind the speech which the speaker is trying to convey. Our main goal in this study is to find emotion behind
the speech using advanceddeep learning algorithms. We have used a popular datasets viz. RAVDESS, TESS, SAVEE to test
out or model.The model is trained in such manner that a set of eight emotions (neutral, calm, happy, sad, angry, fearful,
disgust, surprise) which can be classified by the aforementioned model.Our best and worst results are obtained by Resnet i.e.
82% and MobileNet i.e. 47%, respectively. In order for further research we have created a github repository which can be
accessed via https://github.com/prakharnarayan/SER-with-pre-trained-deep-learning-models
Author - Prakhar A Narayan, L. N. Das, Rakshit Singh, Siddharth Dhawan
Published : Volume-8,Issue-7 ( Jul, 2021 )
DOIONLINE Number - IJAECS-IRAJ-DOIONLINE-18105
View Here
|
|
| |
|
PDF |
| |
Viewed - 41 |
| |
Published on 2021-10-28 |
|
|
|
|
|
|