Applications of hidden Markov model and support vector machine for state estimation / Md.Fayeem Aziz

• Main Author: Md. Fayeem , Aziz
• Format: Thesis
• Published: 2014
• Subjects: T Technology (General); TJ Mechanical engineering and machinery
• Online Access: http://studentsrepo.um.edu.my/8297/2/Final_verson_thesis.pdf; http://studentsrepo.um.edu.my/8297/

This thesis introduces applications of support vector machine (SVM) and hidden Markov model (HMM) for signal processing and image processing. The result of the SVM classifier treated as is used as observation to the HMM and the state is estimated by probabilistic argument maximization. The probability of state is calculated by the classification outcome and the previous state. This method is tested on two case studies. The first case study is about controlling an automated wheel chair using electrooculography (EOG) traces in electroencephalograph (EEG). EOG traces originate from eyeball and eyelid movements and they are embedded in EEG signals collected from the scalp of the user at three different locations. Features extracted from the EOG traces are used to determine whether the eyes are open or closed, and whether the eyes are gazing to the right, center or left. These features are utilized as inputs to a few SVM classifiers whose outputs are regarded as observations to an HMM. The HMM determines the state of the system and generates commands for navigating the wheelchair accordingly. The second case is related to bin level classification and collection scheduling. First the exact bin location and orientation are detected using Hough line detection and angle measurement. Then the Gabor filter (GF) features are extracted from the bin opening in the image and used as inputs to an SVM classifier. The output is the exact bin locations and waste level classification, which is empty, low, full or overflow. The classes of waste level are considered as observation of HMM to estimate the interval to collection time. The system achieves 98% accuracy in estimating the wheelchair navigation command from EOG tracing in EEG signal and 100% accuracy in estimating the waste collection schedule.