Looking here you will see the location and structure of the V4 region

http://www.nature.com/nrmicro/journal/v12/n9/fig_tab/nrmicro3330_F1.html

Then Import Silva db in ARB, open the alignment editor and there you should be able to recover that region. At the moment my ARB is not working, so I cannot give u more details. But maybe you could give it a try and figure it out by your self the right way of doing it.

BBMap contains tools for this purpose, based on primer sequences. Let's say you want the region between primers "ACGT" and "CCCC":

msa.sh in=16s.fa out=x.sam literal=ACGT
msa.sh in=16s.fa out=y.sam literal=CCCC
cutprimers.sh in=16s.fa out=V4.fa sam1=x.sam sam2=y.sam

If you have multiple primers, use a comma-delimited list such as literal=ACGT,ACTT.

I've recently developed a simple Python tool to extract 16S regions with respect to the reference E. coli notation (V4 is mapped from ~515 to ~810). The tool is openly available here https://github.com/grayfall/hypervar. You can install it using pip: pip install git+https://github.com/grayfall/hypervar (I recommend doing it in a virtualenv environment). The installation script adds extract.py executable to your PATH, hence it can be invoked as extract.py -f your_seqs.fna -o extracted.fna -s 515 -e 810 -j cores. I strongly recommend using multiple cores if you are willing to process many thousands of sequences, because the tool uses relatively slow BioPython alignment functions.