Dual-tree Complex Wavelet Support Vector Machine

In this paper, we propose a new support vector machine (SVM) kernel using the dual-tree complex wavelet transform. We have proven that this kernel is an admissible support vector kernel. The reason why we use the dual-tree complex wavelet transform is because it satisfies the approximate shift invariant property, which is very important in signal processing. Experiments on signal regression show that this method is comparable to or even better than the existing SVM function regression with the scalar wavelet kernel, the Gaussian kernel, and the exponential radial basis function kernel. We can apply this SVM kernel to other practical applications as well, e.g., pattern recognition.