Comments on "Canonical coordinates and the geometry of inference, rate, and capacity"
Rae‐Hong Park · IEEE Transactions on Signal Processing · 2002
In a paper by Scharf and Mullis (see ibid., vol.48, p.824-31, Mar. 2000), the circulant Gaussian channel was presented as an example to compute canonical correlations and to derive Shannon's capacity theorem. The objective of this article is to present a simpler real-valued discrete Hartley transform (DHT) representation for a real, symmetric, circulant covariance matrix in place of the complex-valued discrete Fourier transform (DFT) representation.