Minyatür eylemsizlik duyucuları kullanılarak insan hareketlerinin sınıflandırılması

This thesis provides a comparative study on activity recognition using miniatureinertial sensors (gyroscopes and accelerometers) and magnetometers wornon the human body. The classification methods used and compared in thisstudy are: a rule-based algorithm (RBA) or decision tree, least-squares method(LSM), k-nearest neighbor algorithm (k-NN), dynamic time warping (DTW-1 and DTW-2), and support vector machines (SVM). In the first part of thisstudy, eight different leg motions are classified using only two single-axis gyroscopes.In the second part, human activities are classified using five sensor unitsworn on different parts of the body. Each sensor unit comprises a tri-axial gyroscope,a tri-axial accelerometer and a tri-axial magnetometer. Different featuresets extracted from the raw sensor data and these are used in the classificationprocess. A number of feature extraction and reduction techniques (principalcomponent analysis) as well as different cross-validation techniques have beenimplemented and compared. A performance comparison of these classificationmethods is provided in terms of their correct differentiation rates, confusion matrices,pre-processing and training times and classification times. Among theclassification techniques we have considered and implemented, SVM, in general,gives the highest correct differentiation rate, followed by k-NN. The classificationtime for RBA is the shortest, followed by SVM or LSM, k-NN or DTW-1,and DTW-2 methods. SVM requires the longest training time, whereas DTW-2takes the longest amount of classification time. Although there is not a significantdifference between the correct differentiation rates obtained by different cross-validationtechniques, repeated random sub-sampling uses the shortest amountof classification time, whereas leave-one-out requires the longest.