#kernel-trick

One-dimensional data alternates between clusters of + and − points along the $x$ axis, so it is not linearly separable. A student separates it with an SVM using the kernel $K(x_1, x_2) = \cos\tfrac{\pi x_1}{4} \cos\tfrac{\pi x_2}{4} + \sin\tfrac{\pi x_1}{4} \sin\tfrac{\pi x_2}{4}.$ Which feature map $\Phi$ satisfies $K(x_1, x_2) = \Phi(x_1) \cdot \Phi(x_2)$?