International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 06 | Jun 2025
p-ISSN: 2395-0072
www.irjet.net
FACE EMOTION RECOGNITION AND ANALYSIS Robin Nadar1, Saveena Nadar2 M.Tech. Student, Electronics and Telecommunication, SunRise University, Alwar, Rajasthan, INDIA. 2M.Sc. IT, Information Technology, Nagindas Khandwala College, Malad, Mumbai, INDIA. ---------------------------------------------------------------------***--------------------------------------------------------------------(CNNs) for automatic hierarchical feature extraction. The Abstract - The Face Emotion Recognition and Analysis 1
model is trained on the FER-2013 dataset, a well-established benchmark in the domain, which enables the recognition of seven key emotions: happy, sad, anger, surprise, fear, disgust, and neutral. The CNN architecture employed in this work is engineered with multiple convolutional and pooling layers, dropout regularization, and optimized using the Adam algorithm, ensuring both depth and generalization.
system is a real-time application designed to detect human emotions from facial expressions using deep learning. It leverages the FER-2013 dataset and a Convolutional Neural Network (CNN) architecture with 13 layers for accurate classification of seven emotion categories: happy, sad, angry, fear, surprise, disgust, and neutral. A unique feature of the system is its ability to log recognized emotions with timestamps and categorize them into specific time slots (morning, afternoon, evening, night), enabling insightful trend analysis through visualization. The system is implemented using Python, TensorFlow, and Streamlit, achieving an accuracy of 74.62% over 50 epochs. Applications span across healthcare, security, customer service, and user experience evaluation.
A distinct engineering enhancement introduced in this system is the temporal emotion logging mechanism, which records each recognized emotion with a timestamp and categorizes it into defined time slots, morning, afternoon, evening, and night. This time-aware design allows behavioural trend analysis, providing an additional dimension of emotion interpretation not present in most existing models. The logged data is further visualized using dynamic bar graphs and pie charts, enabling both qualitative and quantitative analysis.
Key Words: Face, Emotion, Recognition, Analysis, CNN, Convolutional Neural Network, Neural Network, FER, Face Emotion Recognition and Analysis.
Another critical advancement is the system’s ability to function with low-latency on resource-constrained environments using optimized libraries such as TensorFlow and OpenCV, making it deployable on embedded platforms or edge devices (e.g., Raspberry Pi, Jetson Nano). Unlike conventional FER systems that operate only on single face detection or require cloud backends, this system performs multi-face real-time detection and emotion prediction directly on the host machine, thereby preserving privacy and reducing processing delay.
1. INTRODUCTION The integration of artificial intelligence (AI) into behavioural analysis has revolutionized the way machines perceive and respond to human emotions. Among various biometric indicators, facial expressions are considered the most intuitive and immediate means of non-verbal communication. In recent years, facial emotion recognition (FER) systems have evolved significantly with the advancement of deep learning techniques and the availability of large-scale annotated datasets.
The proposed model is also integrated with a Graphical User Interface (GUI) built using Streamlit, allowing interactive visualization and user engagement without the need for programming knowledge. All detections are stored in a structured CSV format, and the interface includes easy-tounderstand analytics, making it accessible to both engineers and non-technical users.
From an engineering perspective, facial emotion recognition presents a complex, multi-stage problem involving image acquisition, pre-processing, feature extraction, classification, and post-processing. Each stage requires robust algorithmic frameworks and optimized implementations to ensure system accuracy, speed, and real-time operability. Traditional emotion detection systems often relied on handcrafted features (like Local Binary Patterns or Gabor filters), shallow classifiers (like SVM or k-NN), and offline static image analysis. These approaches suffered from limitations such as sensitivity to lighting conditions, poor generalization across diverse populations, and lack of temporal or contextual emotion tracking.
In summary, this research aims to build a technically sound, scalable, and user-friendly emotion recognition framework. The system not only outperforms traditional approaches in terms of real-time capability, contextual analysis, and engineering efficiency, but also opens new possibilities in fields such as adaptive learning, mental health monitoring, smart surveillance, and emotion-aware AI agents.
To address these challenges, this study proposes a real-time, deep learning-based Face Emotion Recognition and Analysis system that leverages Convolutional Neural Networks
© 2025, IRJET
|
Impact Factor value: 8.315
|
ISO 9001:2008 Certified Journal
|
Page 343