Anji Reddy Vaka; G. Niveditha; Anirban Das; Kasiprasad Mannepalli; Rajesh Nekkanti; Jamunadevi C.

BlinkFusion: A Detector-Agnostic, EAR-Interpretable, Real-Time Blink Analysis Pipeline (YOLOv5 + Haar Fallback)

PDF (1632KB), PP.234-257

Views: 0 Downloads: 0

Author(s)

Anji Reddy Vaka ^1,* G. Niveditha ² Anirban Das ³ Kasiprasad Mannepalli ⁴ Rajesh Nekkanti ⁵ Jamunadevi C. ⁶

1. Department of CSE, Lendi Institute of Engineering and Technology, Vizianagaram, Andhra Pradesh 535005, India

2. Department of CSE, Geethanjali College of Engineering and Technology, Medchal, Telangana 501301, India

3. Department of Basic Science and Humanities, College of Engineering and Management, Kolaghat, Purba Medinipur, West Bengal 721171, India

4. Department of ECE, Koneru Lakshmaiah Education Foundation, Vaddeswaram, Guntur(dist), Vijayawada 522502 , India

5. Department of Electrical and Electronics Engineering, Aditya University, Surampalem, East Godavari, Andhra Pradesh 533437, India

6. Department of Computer Technology-PG, Kongu Engineering College, Perundurai, Erode 638060, Tamil Nadu, India

* Corresponding author.

DOI: https://doi.org/10.5815/ijigsp.2026.05.12

Received: 9 Mar. 2026 / Revised: 31 Mar. 2026 / Accepted: 15 Apr. 2026 / Published: 8 Oct. 2026

Index Terms

Blink Fusion, Event Analysis, Eye Aspect Ratio, Hysteresis State Machine, Platt-Calibrated Confidence Score.

Abstract

This paper introduces BlinkFusion, a real-time, interpretable, and detector-independent pipeline for analyzing blink events. The system finds eye ROIs using YOLOv5 as the main detector and a Haar cascade as a backup that has been calibrated. It then stabilizes the detections using lightweight tracking. A small landmark regressor inside each ROI gives six points to calculate the Eye Aspect Ratio (EAR), which keeps the geometric meaning. An uncertainty-weighted smoother combines pose, landmark, and detector confidence. An online hysteresis state machine with minimum-duration and derivative gates makes blink onsets and offsets. Platt-calibrated and fused detector confidences make strong arbitration possible at a reasonable cost. Performance is assessed on EyeBlinkDB (RGB ≥25 fps, ~720p), using subject-independent 10-fold splits and temporal-IoU matching. The model achieves Precision 0.928 ± 0.008, Recall 0.945 ± 0.008, F1 0.936 ± 0.008, AP@tIoU=0.3 = 0.962 ± 0.006, with timing precision of 28.1 ± 1.7 ms onset MAE and 37.1 ± 2.3 ms offset MAE. Condition stratification validates robustness: The values for Bright/Dim F1 are 0.947/0.922, for Frontal/±15°/≥30° yaw F1 they are 0.952/0.936/0.903, for Glasses (No/Yes) they are 0.946/0.925, and for Occlusion (No/Yes) they are 0.948/0.906. With adaptive scheduling (YOLO invocation rate ρ≈0.045), pipeline runs at about 32 FPS on the CPU (about 90 FPS on the observed frame loop) with a fixed 3-frame delay, which is fast enough for real-time use on cheap hardware. Ablations demonstrate progressive improvements resulting from calibrated fusion, tracking, quality-weighted EMA, and hysteresis (F1: 0.903 → 0.952). The approach is modular (you may switch the detector), doesn't need bounding boxes, and makes judgments that can be explained using EAR traces and thresholds. So, BlinkFusion is a useful, ready-to-use solution for HCI and clinical contexts that need clear, accurate, and quick blink analytics.

Cite This Paper

Anji Reddy Vaka, G. Niveditha, Anirban Das, Kasiprasad Mannepalli, Rajesh Nekkanti, Jamunadevi C., "BlinkFusion: A Detector-Agnostic, EAR-Interpretable, Real-Time Blink Analysis Pipeline (YOLOv5 + Haar Fallback)", International Journal of Image, Graphics and Signal Processing(IJIGSP), Vol.18, No.5, pp. 234-257, 2026. DOI:10.5815/ijigsp.2026.05.12

Reference

[1]A. S. Mustafa, R. Sujatha, P. Jessie, R. Mithuna, R. Mekala, & A. Vijay, “A Python Based Machine Learning Framework for Eye Blink Detection with BlueEye Technology.” In 2024 International Conference on IoT Based Control Networks and Intelligent Systems (ICICNIS), pp. 1516-1521, 2024, December. IEEE.
[2]R. Achary, & C. J. Shelke, “E-Authentication System using Eye-Blinking and Deep Learning.” In 2024 Second International Conference on Networks, Multimedia and Information Technology (NMITCON) pp. 1-7, 2024, August.IEEE.
[3]R. Iijima, T. Takehisa, T.Ohki, & T. Mori, “The catcher in the eye: recognizing users by their blinks. In Proceedings of the 19th ACM Asia Conference on Computer and Communications Security”, pp. 1739-1752, 2024, July.
[4]L. Arnold, S. Aryal, B. Hong, M. Nitharsan, A. Shah, W. Ahmed,...& D. Piaggio. “A Systematic Literature Review of Eye-Tracking and Machine Learning Methods for Improving Productivity and Reading Abilities.Applied Sciences”, vol.15, no.6, p.3308, 2025.
[5]J. Alguindigue, A. Singh, A. Narayan, & S. Samuel, “Biosignals monitoring for driver drowsiness detection using deep neural networks.” IEEE Access, vol.12, pp.93075-93086, 2024.
[6]J. Hong, J. Shin, J. Choi, & M. Ko, “Robust eye blink detection using dual embedding video vision transformer.” In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pp. 6362-6372, 2024.
[7]S. Sushmitha, H. B. Prasad, & V. M. Sapna,. “Multi-layer Secure Authentication using Morse code and Eye Blink Detection.” In 2025 9th International Symposium on Innovative Approaches in Smart Technologies (ISAS) pp. 1-8, 2025, June.IEEE.
[8]A. Sukumaran, & A. Manoharan, “Multimodal engagement recognition from image traits using deep learning techniques.” IEEE Access, vol. 12, pp.25228-25244, 2024.
[9]K. Aravinda, B. S. Kumar, B. P. Kavin, & A.Thirumalraj, “Traffic sign detection for real-world application using hybrid deep belief network classification. In Advanced Geospatial Practices in Natural Environment Resource Management” pp. 214-233, 2024.IGI Global Scientific Publishing.
[10]H. A. Madni, A. Raza, R. Sehar, N. Thalji, & L. Abualigah, “Novel transfer learning approach for driver drowsiness detection using eye movement behavior.” IEEE Access, vol. 12, pp.64765-64778, 2024.
[11]S. Gupta, D. Thakur, A. Rastogi, R. Mishra, S. Balabantaray, & M. Bikhani, “Exploring EEG Characteristics and Machine Learning Classifiers for Accurate Detection of Eye-Blink Mistakes.In Explainable AI (XAI) for Sustainable Development” pp. 89-107, 2024.Chapman and Hall/CRC.
[12]J. Xiong, W. Dai, Q. Wang, X. Dong, B. Ye, & J. Yang, “A review of deep learning in blink detection.PeerJ Computer Science”, vol.11, p. e2594, 2025.
[13]D.Tebor, E., x Headley, & M. Karakaya, “Impact of blinking on deep learning based iris recognition.” In Proceedings of the 2024 ACM Southeast Conference, pp. 113-120, 2024, April.
[14]R. Daza, A. Morales, J. Fierrez, R. Tolosana, & R. Vera-Rodriguez, “mEBAL2 database and benchmark: Image-based multispectral eyeblink detection.” Pattern Recognition Letters, vol.182, pp.83-89, 2024.
[15]D. Chavarro, & M. Karakaya, “Blink detection for off-angle iris recognition using deep learning.” In SoutheastCon 2024 , pp. 1256-1261, 2024, March.IEEE.
[16]F. Ata, K. Ayturan, F. Hardalaç, & U. Kutbay, “Deep Learning-Based Eye Blink Detection in Video Frames: Performance Analysis of Various Models”. 2024.
[17]B. Liu, S. W. Lye, & Z. B. Zakaria, “An integrated framework for eye tracking-assisted task capability recognition of air traffic controllers with machine learning.”Advanced Engineering Informatics, vol.62, p.102784, 2024.
[18]A. Balaji, D. Balanjali, G. Subbaiah, A. A.,Reddy, & D. Karthik, “Federated Deep Learning for Robust Multi-Modal Biometric Authentication Based on Facial and Eye-Blink Cues. International Journal on Advanced Computer Engineering and Communication Technology,” vol.14, no.1, pp. 17-24, 2025.
[19]G. Nousias, K. K. Delibasis, & G. Labiris, Blink Detection Using 3D Convolutional Neural Architectures and Analysis of Accumulated Frame Predictions. Journal of Imaging, vol.11, no.1, p.27, 2025.
[20]A. K. Dharmireddy, P. Srinivasulu, M. Greeshma, V. S. Babu, M. R. Kumar, & A. S. Rajasekaran, “Driver Drowsiness Detection Using AIoT and Machine Learning Techniques.In AIoT” pp. 253-273, 2025..Auerbach Publications.
[21]S. K. Alimoon, A. Kiran, B. G. Narayana, A. S. Lekha, and G. V. Narayana, “Vision-Based Driver Fatigue Detection Using Eye and Mouth Aspect Ratios with Machine Learning,” International Journal of Advance Scientific Research and Engineering Trends (IJASRET), vol. 9, no. 3, pp. 26–32, 2025.
[22]P. Kielty, C. Ryan, W. Shariff, J. Lemley, & P. Corcoran, “Event-Based Multi-Task Facial Landmark and Blink Detection.” IEEE Access, 2025.
[23]N. Maheswaran, S. Bose, G. Vijayaraj, G., Gokulraj, T. Anitha, & D. Prabhu, “Early Detection of Child Eye Dryness Using IoT and Deep Learning Enabled AI Model.” In 2025 International Conference on Computing and Communication Technologies (ICCCT) pp. 1-6, 2025, April. IEEE.
[24]L. Arya & Y. K. Sharma, “Safeguarding Women: IoT-Enabled Machine Learning Approach for Threat Detection via EEG and Eye Blink Signals. “International Journal of Computer Information Systems and Industrial Management Applications, vol. 16, no.3, pp. 12-12, 2024.
[25]M. Javed, Z. Zhang, F. H. Dahri, & A. A. Laghari, “Real-time deep fake video detection using eye movement analysis with a hybrid deep learning approach. Electronics,” vol. 13,no.15, p.2947, 2024.

International Journal of Image, Graphics and Signal Processing (IJIGSP)