KMITL Expo 2026 LogoKMITL 66th Anniversary Logo

VIDEO-BASED EMOTION DETECTION FROM FACIAL EXPRESSIONS WITH ROBUSTNESS TO PARTIAL OCCLUSION

Abstract

Facial Expression Recognition (FER) has attracted considerable attention in fields such as healthcare, customer service, and behavior analysis. However, challenges remain in developing a robust system capable of adapting to various environments and dynamic situations. In this study, the researchers introduced an Ensemble Learning approach to merge outputs from multiple models trained in specific conditions, allowing the system to retain old information while efficiently learning new data. This technique is advantageous in terms of training time and resource usage, as it reduces the need to retrain a new model entirely when faced with new conditions. Instead, new specialized models can be added to the Ensemble system with minimal resource requirements. The study explores two main approaches to Ensemble Learning: averaging outputs from dedicated models trained under specific scenarios and using Mixture of Experts (MoE), a technique that combines multiple models each specialized in different situations. Experimental results showed that Mixture of Experts (MoE) performs more effectively than the Averaging Ensemble method for emotion classification in all scenarios. The MoE system achieved an average accuracy of 84.41% on the CK+ dataset, 54.20% on Oulu-CASIA, and 61.66% on RAVDESS, surpassing the 71.64%, 44.99%, and 57.60% achieved by Averaging Ensemble in these datasets, respectively. These results demonstrate MoE’s ability to accurately select the model specialized for each specific scenario, enhancing the system’s capacity to handle more complex environments.

Objective

ปัจจุบันการตรวจจับอารมณ์ของมนุษย์ผ่านการแสดงออกทางใบหน้า (Emotion Detection Using Facial Expression) ได้รับความสนใจมากขึ้น เนื่องจากมีการประยุกต์ใช้อย่างแพร่หลายในหลายด้าน เช่น สุขภาพจิตการศึกษา และการบริการลูกค้า อย่างไรก็ตาม การพัฒนาระบบที่มีความแม่นยำและสามารถทนทานต่อการเปลี่ยนแปลงของสภาพแวดล้อม เช่น การบดบังบางส่วนของใบหน้า หรือสภาพแสงที่ไม่สม่ำเสมอ ยังคงเป็นความท้าทายหลัก โดยเฉพาะการพัฒนาโมเดลที่สามารถทำงานได้ในสภาพแวดล้อมที่หลากหลาย จากการศึกษางานวิจัยเกี่ยวกับ Facial Expression Recognition (FER) ผู้วิจัยพบว่าเทคนิค Frame Attention Network (FAN) ซึ่งเป็นการประยุกต์ใช้กลไก Attention จากงานด้านการประมวลผลภาษาสามารถนำมาใช้เพื่อให้ความสำคัญกับเฟรมที่มีความหมายในวิดีโอ ทำให้ระบบสามารถโฟกัสเฉพาะเฟรมที่แสดงอารมณ์ที่สำคัญได้ ซึ่งส่งผลต่อประสิทธิภาพของโมเดล ทีมวิจัยจึงนำเทคนิคนี้มาปรับปรุงเพื่อเพิ่มความทนทานของระบบในการจัดการสถานการณ์ดังกล่าว เพื่อพัฒนาประสิทธิภาพของระบบให้ดียิ่งขึ้น ทีมวิจัยได้ใช้แนวทาง Ensemble Learning ซึ่งเป็นการรวมผลลัพธ์จากหลายโมเดลที่ถูกฝึกในเงื่อนไขเฉพาะ การใช้ Ensemble ช่วยลดข้อผิดพลาดจากการใช้โมเดลเดียว และเพิ่มความแม่นยำและความน่าเชื่อถือของผลลัพธ์ โดยเฉพาะอย่างยิ่งในสภาพแวดล้อมที่หลากหลาย อย่างไรก็ตาม ทีมวิจัยยังได้ขยายการพัฒนาเพิ่มเติมโดยใช้เทคนิค Multi-Task Learning (MTL) เพื่อให้ระบบสามารถเรียนรู้จากหลายงานพร้อมกัน ซึ่งในงานวิจัยนี้ได้นำ MTL มาใช้ใน Mixture of Experts โดยให้ MTL ทำหน้าที่เป็นกลไก Gating ช่วยเลือกโมเดลที่เหมาะสมกับแต่ละสถานการณ์ เช่น การบดบังใบหน้า ทำให้ระบบสามารถตัดสินใจได้อย่างมีประสิทธิภาพว่าควรใช้โมเดลใดในสภาวะแวดล้อมที่ต่างไป สามารถรักษาความแม่นยำแม้ในสภาวะที่มีความหลากหลายและยังคงรักษาข้อดีในเรื่องของความสามารถในการขยายขนาด (Scaling Up) ได้อย่างมีประสิทธิภาพ

Other Innovations

3D Soundscape Healing: The L-R Beat Exploration in Binaural Beats Therapy

วิทยาลัยวิศวกรรมสังคีต

3D Soundscape Healing: The L-R Beat Exploration in Binaural Beats Therapy

This project explores the therapeutic potential of binaural beats within a 3D soundscape environment, focusing on the effects of left-right (L-R) beating sound positioning. Using Dolby Atmos technology to create immersive auditory experiences, the research aims to investigate how varying spatial beating sound placements in binaural beat therapy influence mental and emotional healing. Binaural beats, a form of auditory brainwave entrainment, have been shown to promote relaxation, reduce anxiety, and enhance cognitive performance. However, there has been limited exploration of how spatial sound technologies, like Dolby Atmos, can enhance the efficacy of these therapies. This study examines how different beating L-R configurations in a 3D soundscape impact the listener’s perception and therapeutic outcomes. Participants will experience binaural beat sessions in various beating L-R orientations, and physiological and psychological measures, such as heart rate variability and self-reported relaxation levels, will be assessed. The results are expected to provide new insights into the interaction between spatial audio environments and sound-based therapies, potentially improving sound therapy practices by leveraging advanced audio technologies.

Read more
Automatic License Plate Recognition Service

คณะวิศวกรรมศาสตร์

Automatic License Plate Recognition Service

This project focuses on the development of an automatic license plate recognition system that supports both standard and special license plates in Thailand. By utilizing Machine Learning technology, the system enhances the efficiency of license plate reading. It can process data from both images and videos. Users can register and subscribe to the service, allowing them to send data for processing through RESTful API, WebSocket, and registered IP cameras.

Read more
13th Celebration of Silk  Thai Silk Road to the World 2024

คณะครุศาสตร์อุตสาหกรรมและเทคโนโลยี

13th Celebration of Silk Thai Silk Road to the World 2024

Industrial Education and Technology, King Mongkut's Institute of Technology Ladkrabang has a vision for sustainable excellence. The mission is to develop learners to be ready for the digital world, develop educational innovations using research as a base, strategic management with good governance, and academic services that benefit society. In this activity, the group of students joined with the Embassy of the Russian Federation in the Kingdom of Thailand, the working group of the Thai Silk and Culture Promotion Association, and the working group of the National Research Office (NRCT) to integrate knowledge to design a silk outfit that combines Thai and Russian cultures, create a network of cooperation in arts, culture, technology, innovation, and dissemination of knowledge and the beauty of Thai silk. The objective is to develop the potential of teachers and students in creative design, listen to the work guidelines from the working group of the Thai Silk and Culture Promotion Association, the working group of the National Research Office (NRCT) via an online meeting. The team of teachers and students from King Mongkut's Institute of Technology Ladkrabang under the name of the "Love Silk" group designed a Thai silk outfit that combines cultures with the identity of Thai silk and studying traditions and cultures of the Russian Federation's clothing. They studied related literature and research documents and integrated knowledge into the design process, inspired by the concept of the Rajapataen outfit. Since the reign of King Chulalongkorn (Rama V) in 1872, together with the clothing culture of the Russian Federation, emphasizing Thai silk, this concept has gone through the process of creating and selecting the design concept (Concept Generation and Selection). The concept received from the embassy was first submitted for feedback on July 25, 2024. There was a suggestion to add more uniqueness to Thai silk through a fashion show presentation by the wife and grandson of the Ambassador of the Russian Federation. Therefore, the designed outfits are 1 set of women's clothes and 1 set of boys' clothes. The women's set has an inner shirt adapted from the royal outfit using silk fabric, Kon Ka-ed pattern, with two separate pieces: 1 shirt and 1 skirt. The jacket is a modified long suit style, plain silk, dark pink and red. The boys' set has a long-sleeved shirt adapted from the contemporary royal style, tailored with cream silk, long slacks tailored with raw betel silk, and a collarless coat with blue silk and lotus pattern. Adapted from the Rajapattan suit with a long collar in an international style. On August 2, 2024, the designed suit and the prototype of the raw fabric suit were brought to the ambassador's wife and nephew to try on. On August 30, 2024, the ambassador was met for the 4th time to bring the silk suit that was cut into a real silk suit. The shirt was given to the ambassador's wife and nephew to try on. It was worn to join the fashion show in the 13th "Thai Silk to the World" Silk Festival at the Naval Auditorium, where lecturers and students from the Faculty of Industrial Education and Technology, King Mongkut's Institute of Technology Ladkrabang, joined the fashion show in the finale round, which proceeded smoothly. After the event, the working group of the Thai Silk and Culture Promotion Association brought the clothes designed and tailored by the "Rak Prae Mai" team to exhibit Thai Silk to the World Exhibition from September 1-8, 2024 at the Emsphere Shopping Mall. The team summarized the report and compiled a complete report. In the implementation of this project, the Faculty of Industrial Education and Technology, King Mongkut's Institute of Technology Ladkrabang has received support and facilitation throughout the project. The budget support from the National Research Council of Thailand (NRCT), the support of fabrics for sewing from the Thai Silk and Culture Promotion Association, and the information for designing valuable silk dresses from the Embassy of the Russian Federation in the Kingdom of Thailand are very important factors that made this operation a success. It is a very important experience for the team of teachers and students of Industrial Education and Technology, King Mongkut's Institute of Technology Ladkrabang. We sincerely hope to receive good cooperation in the future.

Read more