13 12-06-20MultirateSigProcPolyphase
13 12-06-20MultirateSigProcPolyphase
13 12-06-20MultirateSigProcPolyphase
Polyphase Representation
Last time we saw how to obtain the polyphase
representation for the filtering and
downsampling operation of 1 filter. We now
extend this formulation for a bank of N filters.
Here we not only have 1 filter, but N filters, and
hence can assemble the N filter outputs or N
subbands into a vector of N elements,
We also now have N filters instead of 1 filter,
and we assemble the polyphase vectors of our
filters into a matrix, in which each column
corresponds to an anlalysis filter,
x mN n=
k =0 m' =0
y k mm ' g k m ' N n
z = Y k z G k z
X
k =0
with
Perfect Reconstruction
Perfect reconstruction (PR) is defined as a
reconstructed signal which is identical to the
original signal except for a delay,
x n= x nnd
with some delay n d at the original sampling
rate. This delay usually results from our
filtering and the downsampling and upsampling
operations. To obtain PR we can simply take a
look at the output of our synthesis filter bank,
The structure of the polyphase analysis and
synthesis filter bank can be seen also in the
following structure,