The automatic recognition of human actions in video is of great interest in many applications such as automated surveillance, content-based video summarization, video search, and indexing. The problem is challenging due to a wide range of variations among the motion pattern of a given action such as walking across different subjects and the low variations among similar motions such as running and jogging. This thesis has three contributions in a discriminative bottom-up framework to improve the multi-resolution analysis of the motion patterns in video for better recognition of human actions. The first contribution of this thesis is the introduction of a novel approach for a robust local motion feature detection in video. To this end, four different multi-resolution temporally causal and asymmetric filters of log Gaussian, scale-derivative Gaussian, Poisson, and asymmetric sinc are introduced. The performance of these filters is compared with the widely used multi-resolution Gabor filter in a common framework for detection of local salient motions. The features obtained from the asymmetric filtering are more precise and more robust under geometric deformations such as view change or affine transformations. Moreover, they provide higher classification accuracy when they are used with a standard bag-of-words representation of actions and a single discriminative classifier. The experimental results show that the asymmetric sinc performs the best. The Poisson and the scale-derivative Gaussian perform better than log Gaussian and that better than the symmetric temporal Gabor filter. The second contribution of this thesis is the introduction of an efficient action representation. The observation is that the salient features at different spatial and temporal scales characterize different motion information. A multi-resolution analysis of the motion characteristic should be representative of different actions. A multi-resolution action signature provides a more discriminative video representation. The third contribution of this thesis is on the classification of different human actions. To this end, an ensemble of classifiers in a multiple classifier systems (MCS) framework with a parallel topology is utilized. This framework can fully benefit from the multi-resolution characteristics of the motion patterns in the human actions. The classification combination concept of the MCS has been then extended to address two problems in the configuration setting of a recognition framework, namely the choice of distance metric for comparing the action representations and the size of the codebook by which an action is represented. This implication of MCS at multiple stages of the recognition pipeline provides a multi-stage MCS framework which outperforms the existing methods which use a single classifier. Based on the experimental results of the local feature detection and the action classification, the multi-stage MCS framework, which uses the multi-scale features obtained from the temporal asymmetric sinc filtering, is recommended for the task of human action recognition in video.

No description available.

No description available.

No description available.

<p>The analysis of the health condition in Rheumatoid Arthritis (RA) remains a qualitative process dependent on visual inspection by a clinician. Fully automatic techniques that can accurately classify the health of the muscle have yet to be developed. The intended purpose of this work is to develop a novel spatio-temporal technique to assist in a rehabilitation program framework, by identifying motion features inherited in the muscles in order to classify them as either healthy or diseased. Experiments are based on ultrasound image sequences during which the muscles were undergoing contraction. The proposed system uses an optical flow technique to estimate the velocity of contraction. Analyzing and manipulating the velocity vectors reveal valuable information which encourages the extraction of motion features to discriminate the healthy against the sick. Experimental results for classification prove helpful in essential developments of therapy processes and the performance of the system has been validated by the cross-validation technique “leave-one-out”. The method leads to an analytical description of both the global and local muscle’s features in a way which enables the derivation of an appropriate strategy for classification. To our knowledge this is the first reported spatio-temporal method developed and evaluated for RA assessment. In addition, the progress of physical therapy to improve strength of muscles in RA patients has also been evaluated by the features used for classification.</p>

The analysis of the health condition in Rheumatoid Arthritis (RA) remains a qualitative process dependent on visual inspection by a clinician. Fully automatic techniques that can accurately classify the health of the muscle have yet to be developed. The intended purpose of this work is to develop a novel spatio-temporal technique to assist in a rehabilitation program framework, by identifying motion features inherited in the muscles in order to classify them as either healthy or diseased. Experiments are based on ultrasound image sequences during which the muscles were undergoing contraction. The proposed system uses an optical flow technique to estimate the velocity of contraction. Analyzing and manipulating the velocity vectors reveal valuable information which encourages the extraction of motion features to discriminate the healthy against the sick. Experimental results for classification prove helpful in essential developments of therapy processes and the performance of the system has been validated by the cross-validation technique “leave-one-out”. The method leads to an analytical description of both the global and local muscle’s features in a way which enables the derivation of an appropriate strategy for classification. To our knowledge this is the first reported spatio-temporal method developed and evaluated for RA assessment. In addition, the progress of physical therapy to improve strength of muscles in RA patients has also been evaluated by the features used for classification.

Motion and motor patterns of low back pain (LBP) sufferers have been demonstrated as different compared to those without LBP. The mechanism behind such movement alterations is not well known but is believed to be related to pain. Current biomechanical measurement of lumbar curvature and kinematics has limitations for routine clinical use. The aims of this research were to (1) investigate new motion analysis technology for dynamic lumbar curvature (fibre-optic sensors) and higher order kinematic assessment (inertial sensors) within a clinical environment; (2) determine the effect of pain relief on lumbar curvature, kinematics and muscle function in acute low back pain (ALBP) and chronic low back pain (CLBP) sufferers. Dynamic lumbar curvature was found to be reliably measureable in the clinic. Additional analysis demonstrated that regional curvature, as well as sequencing of curvature change measurement was possible. ALBP sufferers display less peak curvature during flexion and lifting compared to CLBP sufferers and both groups demonstrated the greatest curvature change in the second quartile for flexion and lifting and first for extension. Partial pain relief did not increase curvature in either group, and neither group was more likely to respond to pain relief by increasing curvature or altering sequencing. Higher order kinematics could be reliably and readily indentified in the clinic using inertial sensors. Movement-velocity plots were employed to describe the movement trajectory and irregularity. Multivariate analysis of variance revealed that neither partial pain relief (pre vs post) nor chronicity (acute vs chronic) had any effect on lumbar kinematics. Individuals appeared to exhibit different electromyography profiles. Those individuals with little deviation in muscle activation commonly displayed lower kinematic values. Partial pain relief did not alter the pattern of EMG profile, muscle onset timing or peak amplitude. These results suggest that clinicians should not expect automatic alterations in motion and motor patterns following interventions which target and achieve partial pain relief.

No description available.

No description available.

The ability to successfully switch hit, or hit a baseball from both sides of the plate, requires a great amount of practice and coordination bilaterally. This study used three-dimensional kinematic and kinetic data to examine the swing patterns of skilled switch hitters in baseball. Three male minor league and division I collegiate switch hitters participated. Subjects stood on force plates and hit baseballs off a tee while their swings were recorded with a three-dimensional optical motion capture system. Each subject performed twenty total swings, ten from the right and ten from the left. The swings were digitally analyzed and the dependent measures were compared side-to-side. The swing was broken down into specific events and temporal phase parameters were obtained. Peak vertical ground reaction force of each foot and stride length of each swing were also obtained. All variables were statistically analyzed using paired t-tests. The subjects displayed surface swing characteristics side-to-side that appeared identical and statistically there were no significant differences in the swing variables side-to-side. However, each subject had slight internal pattern differences side-to-side that are reported and discussed. Switch hitters are an excellent example of skilled practitioners that can provide insight into questions pertaining to dominance and motor control. Further research is needed with more subjects to explore side-to-side similarities and differences in well-established patterns.