MyFERS: A Convolutional Neural Network (CNN)-based Facial Emotion Recognition System for Mental Health Applications
UNIVERSITI TEKNOLOGI MARA (UITM) SARAWAK BRANCH
About the Project
Mental health affects one in five people worldwide, yet objectively tracking a patient's emotional state in therapy remains difficult. Therapists face cognitive overload from simultaneous observation, dialogue, and note-taking, while standard tools, the clinician-rated HAMD scale and K-10, remain subjective and recall-biased. Existing Facial Emotion Recognition (FER) systems compound this: real-time alerts disrupt clinical dialogue, and foreign-trained models show bias on Southeast Asian faces, leaving no structured post-session record. MyFERS solves this as an offline, CNN-powered desktop platform that turns facial expression into an objective clinical biomarker. During sessions it passively captures frames every 5 seconds, distraction-free, processing each through a fine-tuned ResNet50V2 model, selected after benchmarking against MobileNetV2 and EfficientNet-B0, and refined via two-phase transfer learning from FER2013 to a local Malaysian dataset. Therapists retain full authority through human-in-the-loop correction, while the system auto-generates an Emotional Trajectory Report mapping dominant emotions and affective transitions for post-session reflection. Validated through real-world testing, MyFERS demonstrated credible, reproducible performance as a benchmark for localized clinical FER. Its impact spans therapists, who gain fatigue-free documentation, patients, who gain sensor-free progress tracking, and institutions, which gain encrypted, offline, PDPA-compliant auditing, while advancing UN SDG 3 for mental healthcare access, SDG 9 for sovereign homegrown AI, and SDG 10 for reduced algorithmic bias. Commercially, MyFERS targets Malaysian psychiatric clinics and counselling centres via a one-time licensing model, with a roadmap from pilot deployment to MOH-aligned Clinical Decision-Support certification and nationwide EHR integration, copyright-protected under MyIPO as an exportable ASEAN health-tech solution.
More from the same competition
Browse ShowcasePHARMACOLAB: EMPOWERING MEDICAL STUDENTS THROUGH AN INNOVATIVE DIGITAL APPROACH TO PHARMACOKINETICS LEARNING
JABATAN ANATOMI MANUSIA, FAKULTI PERUBATAN DAN SAINS KESIHATAN
KL3IS-26e-CARE (Electronic Counselling & Assessment for Resilient Employees)
Universiti Malaysia Sarawak (UNIMAS)
KL3IS-263 PHASE WIFI DISTRIBUTION BOX USING TELEGRAM
KOLEJ VOKASIONAL BATU LANCHANG
KL3IS-26