Group delay reduction in FIR digital filters

The causal delay of finite impulse response filters can be minimized by the use of recursive Minimum-phase filters, i.e., IIR filters. However, in IIR filters delay distortion arises which is undesirable. Many algorithms have been proposed, which try to attain both these goals, of minimizing causal delay and delay distortion, simultaneously. Usually for a long input sequence to be filtered, block convolution techniques are used such as overlap-save method (OSM) and overlap-add method (OAM). However, in these methods the output sequence has a finite group delay with respect to input. To reduce that group delay, in this paper, we have proposed the technique of enhanced modified overlap and save method. In our method, we first make the impulse response (IR) causal (i.e., h(n)=0 for n<0) and then circular left shift the impulse response by an amount of (N -1)/2 for N odd and N/2 for N even, where N is the length of the filter. We have used this modified IR in OSM based block convolution technique. Finally, we have defined which samples are to be removed from the final convolution result. This results in our proposed filter which has minimized causal delay and whose group delay is reduced by a factor of 45 as compared to that of the OSM based filter.