Pelacakan Titik Fitur Dengan Kanade-Lucas-Tomasi (KLT) Untuk Pengenalan Ekspresi Mikro Wajah Pada Dataset SAMM

Priska Choirina (Universitas Islam Raden Rahmat, Indonesia)
Ulla Delfana Rosiani (Politeknik Negeri Malang, Indonesia)
Indah Martha Fitriani (Universitas Islam Raden Rahmat, Indonesia)

G-Tech : Jurnal Teknologi Terapan
Vol 8 No 1 (2024): G-Tech, Vol. 8 No. 1 Januari 2024

Micro-expressions are subconscious expressions that occur when a person hides emotions. This research's main contribution is applying the Kanade Lucas Tomasi (KLT) method, which can track motion well. The KLT method utilizes feature points to detect motion in each frame. The observation areas are eyebrows, eyes, and mouth corners with facial markers formed by Discriminative Response Map Fitting (DRMF). The result of tracking all feature points is motion features formed from a vector of the displacement of these points. Feature extraction data is formed with two types: handcrafted data and random sampling. Extracted data is classified by comparing two methods: Support Vector Machine (SVM) and MLP-Backpropagation on the SAMM Dataset. The results of this study show significant results. Random sampling feature extraction data using MLP-Backpropagation produces the highest accuracy of 79.8%. In addition, the processing time on each frame using the proposed method is speedy.