I am working on High Quality HE AAC Decoder for a floating-point DSP.
Both the analylis and synthesis filterbanks are consuming huge cycles.

Filterbanks are as follow:

Analysis

X(k) = ∑ u(n) cos[( PI / 64 ) (2*n-0.5) (k+0.5) ] for n = 0 ... 64 and k = 0 ... 31


Synthesis
Monorate Mode

V(n) = ∑ X(k) cos[( PI / 64 ) (2*n-127.5) (k+0.5) ] for n = 0 ... 64 and k = 0 ... 31

Multirate Mode

V(n) = ∑ X(k) cos[( PI / 128 ) (2n-255) (k+0.5) ] for n = 0 ... 127 and k = 0 ... 64


Please suggest faster algorithms for these filterbanks

Thanks !