Real Time Emotion Detection Using Convolutional Neural Network

Project Title

Real Time Emotion Detection Using Convolutional Neural Network

Project Area of Specialization Artificial IntelligenceProject Summary

Real time emotion detection is an automated system that is developed to detect emotions of human by using convolutional neural network (CNN). In this framework, we take frames from live video and processed it using feature extraction and deep learning technique CNN. To detect the emotion facial attributes extraction by principal component analysis is used for different facial expression with respective emotions. Human faces contain significant information about emotions and mental state of a person. After detecting emotions, it can be further used in many aspects like security, medicines, E-learning and marketing.

Project Objectives

The main objectives of the project are:

• To develop an automatic system capable of detecting human emotions
• To extract the features from a picture or live video
• To identify emotions in real time from various viewing angles 
• To develop a real-time emotion recognition system for detecting the face of human and conclude the detected emotion

Project Implementation Method

In this proposed system, emotions of human are detected by using convolutional neural network (CNN). In this framework, we take frames from live video and processed it using feature extraction and deep learning technique CNN. To study the emotion recognition from the face, a dataset is exploited. The dataset is FER-2013, an open-source dataset created by the P. L. carrier and A. Courvile for project work after that it is publicly shared for Kaggle competition.  This dataset consists of facial images. This dataset consists 35887 number of labelled images. Data consists of 48x48 pixel with 3x3 filters grayscale images of faces. It contains images of 7 facial expressions, with distributions of Angry (4,953), Sad (6,077), Disgust (547), Surprise (4,002), Fear (5,121), Neutral (6,198) and Happy (8,989). From dataset, images are extracted in binary array format and binary value. The whole data is normalized. This data is taken as input and there are seven facial expressions as output. The following methodology steps are used to develop a system:

 • Input Image/Video

• Convolutional Layer

• ReLU

• Max Pooling

• Fully Connected Layer

• Output

Pictorial view of methodology to detect real time emotion

Benefits of the Project

The real time emotion detection system will greatly reduce man power and put the user on ease of dealing with the other humans in different domains and their perspectives. This system can detect the human emotion in Nano seconds by just training the system about the emotions of human using the datasets which contains images of different emotions.

This system can monitor the mood of a student during the live lecture which makes the teachers easy to understand the student’s emotion or mood which makes the teacher to give him/her response accordingly.

Technical Details of Final Deliverable

A smart integrated system which detects real time emotions using CNN.

Final Deliverable of the Project HW/SW integrated systemCore Industry EducationOther IndustriesCore Technology Artificial Intelligence(AI)Other TechnologiesSustainable Development Goals Good Health and Well-Being for People, Industry, Innovation and InfrastructureRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	77398
8gb DDR4 2666Mhz Ram.	Equipment	1	5999	5999
Transcend StoreJet 25C3S USB 3.1 Gen 1 Hard Drive 2TB	Equipment	1	16000	16000
Transcend ESD240C 480 GB Portable SSD	Equipment	1	13499	13499
Canon EOS 450D DSLR Camera with lens	Equipment	1	33900	33900
Printing Cost	Miscellaneous	1	4000	4000
Stationary	Miscellaneous	1	2000	2000
Overheads	Miscellaneous	1	2000	2000