19CS71C- Research Paper and Patent Review

Weekly Report - 13

Title: Emotion Detection Using Deep Learning

Problem Statement:
The challenge of accurately recognizing and interpreting human facial emotions from
various angles and perspectives poses significant difficulties in the fields of medical research
and computer science. Traditional methods in facial emotion recognition (FER) rely on feature
extraction and feature selection using Convolutional Neural Networks (CNNs). However, these
methods often struggle with images that are not captured from a frontal view, limiting their
effectiveness in real-world applications. This limitation is particularly critical for enhancing
patient care and supporting children with speech impairments, where accurate emotion
detection can provide essential psychological insights.The proposed research aims to develop
an advanced facial emotion recognition model utilizing a Bi-Directional Long Short-Term
Memory (Bi-LSTM) network. This model seeks to overcome the limitations of conventional
CNN-based approaches by accurately recognizing facial emotions from images captured at
various angles. The primary objective is to create a robust system that can identify and interpret
a wide range of emotional states, enhancing its application in medical and social contexts.

No. of Papers Reviewed: 2

Paper 1:
Title: Facial Emotion Recognition Using Convolutional Neural Network
Authors: Shrey Modi, Mohammed Husain Bohara
Publisher: IEEE
Date of Publication: 2021
Proposed Work: This paper presents a deep learning-based facial emotion recognition system
using Convolutional Neural Networks (CNN). It aims to detect emotions such as happiness,
sadness, and neutral from static facial images. The authors apply CNNs along with transfer
learning techniques and data augmentation. The model was trained on the FER-2013 dataset,
achieving a higher accuracy compared to traditional methods. The study also compares the
CNN model with transfer learning models and highlights CNN’s better performance after 35
epochs, achieving accuracy over 90%(modi2021).

Paper 2:
Title: Facial Emotion Analysis Using Deep Convolution Neural Network
Authors: Rajesh Kumar G A, Ravi Kant Kumar, Goutam Sanyal
Publisher: ICSPC
Date of Publication: 2017
Proposed Work: This paper discusses emotion detection using deep learning, specifically
CNN, trained on the FERC-2013 and CK+ datasets. The authors employ Viola-Jones for face
detection and train their CNN model on seven basic emotions, such as happy, sad, angry, and
neutral. The results show that the system achieved about 90% accuracy in real-time emotion
detection. The research also focuses on detecting emotions at different intensity levels,
showcasing how the model adapts as emotions shift(kumar2017).
Comparative Analysis:
Content Paper 1 Paper 2
Objective Classify emotions (happy, Classify seven basic
sad, neutral) using CNN emotions using CNN,
with transfer learning focusing on intensity levels
techniques. and real-time emotion
detection.
Dataset Used FERC-2013 and CK+
FER-2013 dataset with
datasets (emotion intensity
35,685 grayscale images
analysis).
(48x48 pixels).
Technique Used CNN with transfer learning
CNN combined with Viola-
and data augmentation to
Jones for face detection and
improve emotion
emotion intensity
classification.
prediction.
Modules Description 1. Preprocessing (grayscale, 1. Face detection using
resizing) 2. Feature Viola-Jones 2. Feature
extraction using CNN 3. extraction using CNN layers
Classification using 3. Prediction of emotion
softmax. intensity.
Performance Measure Achieved accuracy over Achieved about 90%
90% after 35 epochs, with accuracy with real-time
CNN outperforming transfer emotion detection and
learning models. intensity analysis.
Future Work Improve emotion intensity
Extend the model to
prediction and integrate
classify additional emotions
micro-expression analysis
(e.g., anger, fear) and
for more complex real-time
develop real-time
applications.
applications using OpenCV.
Conclusion:
Both papers focus on using CNNs for facial emotion recognition but differ in their
scope. Paper 1 emphasizes using CNNs and transfer learning to improve accuracy in classifying
emotions from static images, achieving high accuracy with the FER-2013 dataset. Paper 2 takes
a broader approach by analyzing emotion intensities using both the FERC-2013 and CK+
datasets, achieving comparable accuracy in real-time emotion detection. Both studies offer
valuable insights into deep learning for emotion recognition, but Paper 2 stands out for its focus
on real-time detection and emotion intensity analysis, making it more applicable for dynamic
environments like healthcare and human-computer interaction.

Project Guide Course Instructor