Fig 1.
Principle comparison of between US, DTW and our method.
Fig 2.
Overview of the proposed algorithm.
Fig 3.
Illustration of similarity distances calculation based on Bayesian fusion.
Fig 4.
Example for Bayesian fusion calculation.
Fig 5.
Retrieval precision comparisons between DTW and CCA-base, class-based and Bayes fusion.
Table 1.
List of 30 randomly selected motion groups (each group includes 10 motions).
Fig 6.
Comparison of retrieval precision of the 30 motion groups.
Fig 7.
PR diagram and matrices comparisons for motion category “pick up box” on the 30 groups.
Fig 8.
The PR diagram comparisons for motion categories on the 30 groups.
Fig 9.
The examples for PR diagram comparisons for motion categories on the whole HDM cut database.
Fig 10.
The PR diagram and matrices comparisons for motion subject “walk” on the entire CMU database.
Fig 11.
The PR diagram for motion subjects on the whole CMU database.
Fig 12.
Examples for motion frames clustering using the FCM in HDM database.
Fig 13.
Confusion matrices for motion groups “sneak2StepsRStart” using SLBF-based motion retrieval method, calculated on the HDM05 cut database using FCM clustering feature set and the precision-recall diagram.