The peaks in the output correspond to locations where there are matches. You can divide the target sequence by its squared norm appropriately before filtering, and then instead of peaks, you just need ...