We　study　the　problem　of　how　to　recognize　a　person　only　by　his　eyebrow　based　on　Hidden　Markov　Models　(HMM).　By　experiments　on　a　small-scale　eyebrow　image　database　taken　from　27　subjects,　we　show　that　our　HMM-based　eyebrow　recognition　method　can　achieve　the　highest　accuracy　of　92.6%,　based　on　the　relation　of　its　accuracy　to　the　number　of　observation　symbols　and　that　of　states.　Hence,　we　conclude　that　human　eyebrow　can　work　as　a　biometric　with　certain　possibility　and　feasibility.