I'm following the derivation in this paper A Comb Filter Design Using Fractional-Sample Delay to obtain the objective function for the least-squares filter design.
N-order FIR filter:
$H(z) = \sum_{n=0}^N h(n)z^{-n}$
Frequency response:
$H(\omega) = \mathbf{h}^T\mathbf{e}(\omega) = \mathbf{e}^T(\omega)\mathbf{h}$
where $\mathbf{h}=[h(0)\quad h(1) \quad\dots \quad h(N)] $ and $\mathbf{e}=[1\quad e^{-j\omega} \quad\dots \quad e^{-jN\omega}] $
Least squares error:
$J(h)=\int_{\omega \in R^+\cup R^-} |H(\omega)-F_d(\omega)|^2 d\omega$
where $R^+=[0,\alpha\omega]$ and $R^-=[-\alpha\omega,0]$
Which is rewritten in the quadratic form: $J(h) =h^TQh - 2h^Tp + c$
$Q$, $p$, and $c$ are given in the paper as follows:
$F_d(\omega)= e^{-jD\omega}$ and $h(n)$ is real.
I've been having trouble with getting these expressions for the Q matrix and p vector. How can I obtain them?