Focusing　on　the　problem　that　peak　frequencies　of　seismic　wavelets　estimated　by　common　methods　are　imprecise,　an　eigenstruct　(ES)　method　for　estimating　peak　frequencies　of　seismic　wavelets　is　proposed,　in　which　the　self-correlation　matrix　of　a　seismic　wavelet　is　eigen-decomposed　firstly,　and　then　the　subspace　of　peak　frequency　component　is　approximated　by　the　eigenvector　corresponding　to　the　maximum　eigen-value,　the　subspace　of　other　frequency　components　is　approximated　by　residual　eigen-vectors,　and　the　ratios　of　projected　coefficients　for　different　sine-waves　on　these　two　subspaces　are　computed,　finally　the　harmonic　frequency　corresponding　to　maximum　ratio　is　the　peak　frequency　of　the　seismic　wavelet.　This　method　can　give　more　precise　peak　frequency　of　a　seismic　wavelet　than　common　methods.　Simulation　results　of　synthetic　zero-offset　vertical　seismic　profiles　show　that　the　precision　of　quality　factors　and　peak　frequencies　estimated　by　ES　method　is　higher　than　that　of　fast　Fourier　transform　and　Burg　maximum　entropy,　and　the　variance　of　quality　factors　decreases　by　about　43%.