Decomposing a matrix as the product of rotations

Question

I'm reading an article about joint diagonalization algorithms. The article states without proof that any nonsingular $n \times n$ matrix $Q$ can be decomposed as \begin{align*} Q = \prod_{1 \leq p < q \leq n} \Theta_{pq} \end{align*} where $\Theta_{pq}$ is itself the product of a Givens rotation matrix and a hyperbolic rotation matrix. Can someone provide a proof of this claim?