M.M. Beigi, M. Wang, and A. Zell (Germany)
Biosonar signal processing, Pattern recognition, Kernels.
We consider the problem of biosonar landmark classifi cation as an example of random and non-stationary sig nal classification in which finding robust and structure in dependent features for classification is not trivial. Time frequency domain studies show that despite the seemingly randomness of those signals, there are local temporal simi larities, independent of the position of occurrence in echoes of each object that reflect the intrinsic similarities between the echoes and also a self similarity in the objects. In this paper we suggest a time resolved spectrum kernel for ex tracting the local similarities (subsequence similarity) in time series in general, and as an example in biosonar sig nals. We implemented this kernel using dynamic program ming and could get accurate results using a low number of echoes needed for training compared with the methods in which finding specific features in each echo were followed.
Important Links:
Go Back
