Title: Dynamic spectrum classification by kernel classifiers with divergence-based kernels and its applications to acoustic signals
Authors: Tsukasa Ishigaki, Tomoyuki Higuchi
Addresses: Center for Service Research, National Institute of Advanced Industrial Science and Technology (AIST), 2-41-6 Aomi, Koto-ku, Tokyo, 135-0064, Japan. ' Department of Statistical Modelling, The Institute of Statistical Mathematics, 4-6-7 Minami-Azabu, Minato-ku, Tokyo, 106-8569, Japan
Abstract: In the kernel method, appropriate selection of the kernel function is important for the construction of a high-performance classifier. The present paper describes a high-accuracy dynamic spectrum classification method using kernel classifiers with a divergence-based kernel. We introduce the divergence, which is a metric between two probability distributions, as a kernel function for similarity calculations of two dynamic spectra with appropriate statistical signal processing. The method is applied to two problems of acoustic signal classification: 1 identification of the condition of hazelnut shells using acoustic signals to maintain the quality and safety of the hazelnut product; 2 detection of worn-out banknotes by using acoustic signals to facilitate identification of counterfeit banknotes. In both applications, classification using the divergence-based kernel demonstrates better performance than classifications using popular kernels such as the Gaussian kernel or the polynomial kernel.
Keywords: kernel classifiers; divergence-based kernels; acoustic signals; spectrum classification; hazelnut shells; hazelnut products; worn-out banknotes; counterfeit banknotes; counterfeit identification.
International Journal of Knowledge Engineering and Soft Data Paradigms, 2009 Vol.1 No.2, pp.173 - 192
Available online: 25 Jan 2009 *Full-text access for editors Access for subscribers Purchase this article Comment on this article