I observe that the majority of researchers in the human following robot use many features such as the color of clothes, height, width, gait, etc.
when they evaluate their approaches first, evaluate every feature separately, then integrate them so that performance is high in an integrating case.
why evaluate them separately whereas that integrating case always is best? please, l need to know the academic purpose of this?
The answer Redwhan is several fold. Firstly, the more features you evaluate the more completely you can specify (that is to say identify) the object or person. However any one of the parameters can achieve the same goal in a specific scenario such as “Chase the man in the red coat” or ”Follow the pattern of movements of a customer in a shop”. Next, you may improve the efficiency of your experiment and analysis by classifying and rating the relative significance of the measurements and find the best subset that enable an acceptable answer to be found at a defined probability of correctness level. Clearly, you may achieve a marginally ‘improved’ answer by summing many parametric results, but the cost may be insupportably high or take far too long.
David Oulton Thank you so much for your answer, it has removed some of ambiguity for me
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Publications