First, generate intervals for hg19 (perhaps stripping out non-nuclear and mitochondrial chromosomes):

$ fetchChromSizes hg19 | awk -v FS="\t" -v OFS="\t" '{ print $1, "0", $2; }' | grep -v "[_*_|MT]" | sort-bed - > hg19.nuc.bed

To calculate coverage:

$ bedmap --skip-unmapped --delim '\t' --echo --bases-uniq --echo-ref-size --bases-uniq-f hg19.nuc.bed <(sort-bed peaks.bed) > coverage.bed

The last three columns of coverage.bed are per-chromosome calculations:

• The number of bases in each hg19 chromosome that have coverage by peaks
• The length of the chromosome
• The fraction of each hg19 chromosome that has coverage by peaks (i.e., bases of coverage, divided by chromosome length)

The order of the specified bedmap operands --bases-uniq, --echo-ref-size, and --bases-uniq-f write column values in that same order.

If you want a whole-genome percentage, you could sum the number of bases, sum the chromosome lengths, and divide the two sums at the end:

$ bedmap --skip-unmapped --delim '\t' --echo --bases-uniq --echo-ref-size hg19.nuc.bed <(sort-bed peaks.bed) | awk -v FS="\t" -v OFS="\t" '{ b += $(NF-1); l += $NF; } END { print b/l; }'

Using a toolkit like bedops lets you do this calculation more easily with more complex inputs. For example, you might want GC content over all of the mappable genome, i.e. everything except for blacklisted regions. This is a bit more difficult without a set-operations based approach.

A solution with awk and the bc (basic calculator):

#/ Get chromsizes hg19:
wget http://hgdownload.soe.ucsc.edu/goldenPath/hg19/bigZips/hg19.chrom.sizes

#/ Dummy BED file:
$ cat test.bed 
chr1    100 100000
chr2    20  2000000

#/ Calculate percent of genome represented by test.bed:
bc <<< "scale=10; 100 * $(awk '{sum=$3-$2}END{print sum}' test.bed) / $(awk '{sum+=$2}END{print sum}' hg19.chrom.sizes)"

Answer here would be .0637512652%