Electroencephalogram　(EEG)　is　a　physiological　signal　used　to　monitor　the　electrical　activity　within　the　cortical　monitoring　area　of　the　patient＇s　brain.　It　is　commonly　used　for　real-time　detection　of　seizure　states　and　to　determine　seizure　types.　Traditional　manual　detection　methods　are　not　only　time-consuming　but　also　subjective.　Therefore,　using　deep　learning　methods　to　automatically　classify　various　types　of　epileptic　signals　can　not　only　reduce　the　long-term　monitoring　work　of　neuroscientists　on　patients＇　EEG　signals,　but　also　improve　the　accuracy　of　patient　diagnosis　to　a　certain　extent.　In　this　study,　we　first　used　transverse　central　parietal　(TCP)　montage　to　emphasize　the　spike　activity　and　reduce　the　influence　of　noise　on　the　publicly　available　Temple　University　Hospital　(TUH)　dataset,　resulting　in　20　channels　of　montage-processed　data　information.　Then,　we　sampled　each　epileptic　seizure　data　34　times　on　all　channels　using　a　1-second　time　window　and　0.25-second　time　step,　and　extracted　the　amplitude　of　1-12Hz　from　each　channel　using　fast　Fourier　transform.　Combined　with　the　proposed　model,　experiments　were　conducted　for　training　and　validation,　achieving　an　accuracy　of　88.38%　in　the　classification　of　seven　different　types　of　epileptic　seizures,　which　outperformed　some　existing　traditional　methods.　This　indicates　that　the　proposed　method　can　be　effectively　used　in　the　task　of　classifying　epileptic　seizures.　　©　2023　IEEE.