Please elaborate on your requirement, and also give us some details on what you've tried.

I tried to write a perl code to subtract 2000bp and add 2000 to the CpG island region to get CpG shore regions (55436 CpG shores) [Note: I removed chrUn, random and hap regions from hg19 genomes]. and then use bedtools subtract to remove CpG shores overlapped with CpG island (48790) and then merge the CpG shores which are overlapped within CpG shore regions (46817 CpG shore). Eventually, 46817 CpG shore region were obtained. and you can download this annotation: hg19.cpgshoreExt.txt

Similar idea, we can obtain genome regions for CpG shelf regions. 93634 regions were obtain from CpG shore files. 44844 remains after remove regions overlapping with CpG island, 41934 regions remained after remove regions overlapping with CpG shores and then 40812 regions remained after merge within CpG shelf. hg19.cpgshelfExt.bed can be downloaded.

#!/usr/bin/perl
use strict;
use Cwd;

# Creat CpG shore region based on CpG island 

my $CGI=@ARGV[0];
print "Reference CGI file is $CGI\n";

open F,$CGI || die "Cannot open $CGI";
open OUT,">hg19.cpgShoreExt.sort.txt" || die "Cannot open $CGI";

while(<F>){
my @line=split /\t/;
my $start=$line[1]-2000;
my $end=$line[2]+2000;
my $ID1="$line[0]:$start-$line[1]";
my $ID2="$line[0]:$line[2]-$end";

my $tmp2=$line[0]\t$stat\t$line[1]\t$ID1\n";
my $tmp2="$line[0]\t$line[2]\t$end\t$ID2\n";

print $tmp1;
print $tmp2;

print OUT $tmp1;
print OUT $tmp2;

}

Check this solution too,