PSF and Weighted Overlap Add

Free Books Spectral Audio Signal Processing

Using ``square-root windows'' $\sqrt{w}$ in the WOLA context, the valid hop sizes are identical to those for in the OLA case. More generally, given any window for use in a WOLA system, it is of interest to determine the hop sizes which yield perfect reconstruction.

Recall that, by the Poisson Summation Formula (PSF),

$\displaystyle \zbox {\underbrace{\sum_m w(n-mR)}_{\hbox{\sc Alias}_R(w)} = \underbrace{\frac{1}{R}\sum_{k=0}^{R-1} W(\omega_k)e^{j\omega_k n}}_{\hbox{\sc DFT}_R^{-1} \left[\hbox{\sc Sample}_{\frac{2\pi}{R}}(W)\right]}} \quad \omega_k \isdef \frac{2\pi k}{R} \protect$

(9.39)

For WOLA, this is easily modified to become

$\displaystyle \zbox {\underbrace{\sum_m w(n-mR)f(n-mR)}_{\hbox{\sc Alias}_R(w\cdot f)} = \underbrace{\frac{1}{R}\sum_{k=0}^{R-1} (W\ast F)(\omega_k)e^{j\omega_k n}}_{\hbox{\sc DFT}_R^{-1} \left[\hbox{\sc Sample}_{\frac{2\pi}{R}}(W\ast F)\right]}} \quad \omega_k \isdef \frac{2\pi k}{R}$

(9.40)

where is the analysis window and is the synthesis window.

When , this becomes

$\displaystyle \underbrace{\sum_m w^2(n-mR)}_{\hbox{\sc Alias}_R(w^2)} = \underbrace{\frac{1}{R}\sum_{k=0}^{R-1} (W\ast W)(\omega_k)e^{j\omega_k n}}_{\hbox{\sc DFT}_R^{-1} \left[\hbox{\sc Sample}_{\frac{2\pi}{R}}(W\ast W)\right]}, \quad \omega_k \isdef \frac{2\pi k}{R}$