Walking pattern classification in children with cerebral palsy: A wavelet network approach

Cerebral palsy is a neuro- musculoskeletal condition which affects human gait and balance. Intelligent diagnosis of this condition can have many advantages in clinical decision making. In this paper a novel classification algorithm is proposed in order to distinguish walking patterns in children with cerebral palsy based on kinetic and acceleration data measurements. This algorithm addresses the significance of wavelet network classifier model in pathologic pattern classification. The proposed algorithm consists of three main steps .In the first step feature extraction is implemented using discrete wavelet transform. The second step is feature selection in order to decrease the input data space and omit redundant features. For this purpose Shannon entropy is recruited in order to select the most relevant features which have the most effects on walking patterns. The final step is developing a wavelet network classifier model in order to classify walking patterns in to four groups: healthy, severe cp condition, moderate cp condition and minimal cp condition. Subject pools consist of 51 healthy children and 48 children with cerebral palsy which are selected to obtain walking patterns in the frame work of force, moment and acceleration components.