Analysis of Pattern Similarity between ChoiWilliams Kernel and BornJordan Kernel Abstract.pdfVIP

Analysis of Pattern Similarity between Choi-Williams Kernel and Born-Jordan Kernel Ming Li, Weijia Jia, *Wei Zhao Dept. of Computer Science, City University of Hongkong, Tat Chee Ave., Kowloon, Hongkong Email: {mingli, wjia}@cs.cityu.edu.hk *Dept. of Computer Science, Texas AM University, State College, USA *Email: zhao@cs.tamu.edu Abstract f (t ) = A cos(n ω t + ϕ ) (4) ∑ n 1 n This paper analyzes the pattern similarity of two types of where, An and ϕ n denote the amplitude and the initial kernels, Sinc kernel (i.e., Born-Jordan’s) and Choi- phase of nth harmonic respectively. Thus, Williams kernel (i.e., Gaussian’s). Correlation coefficients are used to characterize the dissimilarities τ τ τ between the two kernels. The correlation curve in the f (t + 2 )f (t − 2 ) = ∑A n cos[n ω 1 (t + 2 ) +ϕ n ] (5) sense of L2 between Born-Jordan kernel and Choi- τ Williams kernel is carried out. The analysis result shows ∑An cos[nω 1(t − 2 ) +ϕ n ] that the maximum correlation value is less than 0.6. Hence, the curve may help us interpret the differences Because of the product f (t + τ )f (t − τ ) , A (u ,τ )