You have a couple of options:
geom_freqpoly(aes(y = ..count.. / sum(..count..)))
which is probably what you want. Then there's:
geom_freqpoly(aes(y = ..ndensity..))
which is the density estimate, but scaled to range from 0 to 1. (i.e. it will always range from 0 to 1). And finally, the associated:
geom_freqpoly(aes(y = ..ncount..))
which is similar, but for the counts. You can read about the options at ?stat_bin
.