4.3. The Evaluation of Similarity
For motion retrieval, hyper sausage neurons covering in high-dimensional
space is used to estimate the distribution of every motion data in the space. If
a frame’s feature vector in a motion clip have a same mark to pose Pk , it can
be considered as a similar object to Pk and be noted as k in Cover
Sequence, or else it is noted as 0 . Cover Sequence noted as C is an integer
array to record the similar status of a motion sequence, for example in Eq4.4,
and it can be simplified. In sequence, zero elements are removed, and
continuous same elements are denoted by one element.