Adegun, Iyanu Pelumi2018-03-132018-03-132017Adegun, Iyanu Pelumi (2017) Automatic recognition of micro-expressions using local binary patterns on three orthogonal planes and extreme learning machine, University of the Witwatersrand, Johannesburg, https://hdl.handle.net/10539/24181https://hdl.handle.net/10539/24181A dissertation submitted in fullment of the requirements for the degree of Master of Science to the Faculty of Science, University of the Witwatersrand, Johannesburg, September 2017Recognition of micro-expressions is a growing research area as a result of its application in revealing subtle intention of humans especially under high stake situations. Owing to micro-expressions' short duration and low inten- sity, e orts to train humans in their recognition has resulted in very low performance. The use of temporal methods (on image sequences) and static methods (on apex frames) were explored for feature extraction. Supervised machine learning algorithms which include Support Vector Machines (SVM) and Extreme Learning Machines (ELM) were used for the purpose of classi- cation. Extreme learning machines which has the ability to learn fast was compared with SVM which acted as the baseline model. For experimentation, samples from Chinese Academy of Micro-expressions (CASME II) database were used. Results revealed that use of temporal features outperformed the use of static features for micro-expression recognition on both SVM and ELM models. Static and temporal features gave an average testing accuracy of 94.08% and 97.57% respectively for ve classes of micro-expressions us- ing ELM model. Signi cance test carried out on these two average means suggested that temporal features outperformed static features using ELM. Comparison between SVM and ELM learning time also revealed that ELM learns faster than SVM. For the ve selected micro-expression classes, an av- erage training time of 0.3405 seconds was achieved for SVM while an average training time of 0.0409 seconds was achieved for ELM. Hence we can sug- gest that micro-expressions can be recognised successfully by using temporal features and a machine learning algorithm that has a fast learning speed.Online resource (xiii, 68 leaves)enBinary system (Mathematics)Orthogonal polynomialsThesis