The amount of daily physical activity over the lifetime of a person, has a positive impact on his/her overall health and may reduce his/her overall risks of developing diseases. Walking is one of the most common daily physical activities. The assessment of walking patterns plays a key role in energy expenditure estimation and in understanding the relationship between daily physical activity and functional health status in humans. This thesis investigates human movement on inclined terrains while walking in an unconstrained environment. Human gaits are analysed, using a number of novel accelerometry-based features which are used for gait pattern classification. A novel real-time algorithm is then developed for estimating exercise rate. Cross-fold validation demonstrates that the proposed gait feature extraction and classification algorithm achieves 82.46% accuracy, in terms of overall classification, for walking patterns on seven different inclined terrains including level terrains, two different grades of uphill/downhill and upstairs/downstairs. A method of automated gait segmentation of the gait cycle has been developed and the reliability of the segmentation is reported. In addition, an exercise rate estimation algorithm has been proposed and experimental results show that the algorithm is effective in estimating exercise rate for common rhythmical aerobic exercises, such as walking, cycling and rowing. The proposed gait feature extraction and classification algorithm demonstrates that it has good potential for improving the accuracy of daily physical activity energy expenditure estimates, with accurate measures of terrain inclinations.