International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 12 | Dec 2025
p-ISSN: 2395-0072
www.irjet.net
Facial Expression Detection Using Image Processing Akshata G1, Bhumika B2, Praveen K3, Sheetal P4, Prof. Pooja Shindhe5 1,2,3,4 UG Student, Dept. of ECE Engineering, VDIT Haliyal, Karnataka, India 5Assistant Professor, Dept. of ECE Engineering, VDIT Haliyal, Karnataka, India
---------------------------------------------------------------------***--------------------------------------------------------------------Abstract - Facial expressions are one of the simplest and 1. INTRODUCTION most natural ways through which people show their emotions. In recent years, there has been a growing interest in developing systems that can automatically understand these expressions, especially for applications in interactive devices, education, healthcare, and security. However, many of the existing solutions depend on machine learning or deep learning models, which require large datasets, complex training procedures, and powerful hardware. These limitations make them difficult to use in small academic projects or in environments where resources are limited.
Facial expressions are one of the most natural ways through which people communicate their feelings and intentions. Without using a single word, a person can express happiness, sadness, excitement, fear, or neutrality simply through subtle movements of the eyes, eyebrows, and lips. Because of this, facial expression recognition has become an important area of study in fields such as human– computer interaction, security systems, education, gaming, and assistive technologies. In recent years, most facial expression detection systems have shifted toward machine learning and deep learning models. These techniques offer high accuracy but depend heavily on large datasets, long training times, and highperformance hardware. Such requirements make them less suitable for academic projects, real-time low-resource applications, or situations where transparency and simplicity are preferred. For many practical scenarios, a rule-based image-processing approach can still provide a reliable and efficient alternative.
In this work, we present a rule-based facial expression detection system that relies completely on basic imageprocessing techniques rather than trained models. The system focuses on identifying five emotions—Happy, Sad, Neutral, Excited, and Fear—by observing simple and clear visual changes in different parts of the face. The face is first captured and preprocessed using grayscale conversion, smoothing, and contrast adjustment. Instead of using CNN or Haar Cascade, the face region and important areas such as the eyes, eyebrows, and lips are located using proportional measurements.
This project focuses on developing a purely rule-based facial expression detection system that does not use any trained models such as CNN or Haar Cascade. Instead, it relies entirely on classical image-processing methods to identify five major expressions: Happy, Sad, Neutral, Excited, and Fear. The system analyzes characteristic changes in the facial regions—especially the eyes, eyebrows, and lips—to determine the user’s emotion. By applying geometric relationships, pixel-level differences, and threshold-based rules, the system can classify expressions in real time.
From these regions, features like the curve of the lips, the openness of the eyes, and the position of the eyebrows are measured. These measurements are then compared with carefully chosen rules and thresholds to determine the emotion. Once an expression is detected, the system automatically plays a video linked to that emotion, making the output more interactive and user-friendly. The overall approach is simple, transparent, and easy to understand. It does not require any dataset, training, or heavy computation. Testing shows that the system works well under normal lighting and frontal face positions.
The motivation behind this work is to create a simple, lightweight, and interpretable method that performs well even without advanced computational resources. Since the logic is fully rule-based, the system offers clear understanding of how each expression is detected, making it highly suitable for academic research, prototype development, and environments where deep learning cannot be implemented.
Key word: Facial Expression Recognition, Digital Image Processing, Rule-Based Classification, Facial Feature Extraction, Threshold-Based Decision Making
© 2025, IRJET
|
Impact Factor value: 8.315
|
ISO 9001:2008 Certified Journal
|
Page 742