Consequently, we only choose some effective wavelet
coefficient as gait features, which are used as the ANN model
inputs. This allows us to construct a novel wavelet-based ANN
gait classifier, that is, the orthonormal discrete wavelet
transform is firstly used to extract some good gait features from
the original gait data, and then the extracted gait features are
used to initiate the training set of ANN for gait classification.
Therefore, we can assess the change of gait function effectively.