Output of FIR filter is convolution of input signal, x[n] and impulse response of filter, h[n]. Practically length of input signal is very long and and convolution becomes complicated. Therefore for filtering real time input signal there are two methods.
1. Overlap Add Method (OAM) 2. Overlap Save Method (OSM)
In both the methods the input signal is decomposed into small parts. In OAM ,length of decomposed signal is L,where L=N-M+1. Linear convolution of decomposed input signal and h[n] is carried out and final output is found out by shifting and adding individual outputs. In OSM ,input is decomposed by adding some preceding values. Circular convolution is done with h[n] after zero padding h[n].Final output of OSM is obtained after discarding the extra preceded values. Thus, OAM and OSM are block processing techniques.
1. Overlap Add Method (OAM) 2. Overlap Save Method (OSM)
In both the methods the input signal is decomposed into small parts. In OAM ,length of decomposed signal is L,where L=N-M+1. Linear convolution of decomposed input signal and h[n] is carried out and final output is found out by shifting and adding individual outputs. In OSM ,input is decomposed by adding some preceding values. Circular convolution is done with h[n] after zero padding h[n].Final output of OSM is obtained after discarding the extra preceded values. Thus, OAM and OSM are block processing techniques.
In OSM you have mentioned decomposition is done by adding 'some' preceding values. Cam you please elaborate ?
ReplyDeleteThe last 3 or 4 values of previous decomposed signal are added to next decomposed signal.
DeleteOSM and OAM methods are useful for long sequence input signal.
ReplyDeletecomputationally both OSM and OAM are of equally beneficial.
DeleteProcessing is faster because of use of FFT for OAM and OSM
ReplyDeleteUseful for processing long streams of input
ReplyDeleteOSM and OAM are computationally same
ReplyDelete