Functional　electrical　stimulation　(FES)　stimulates　the　patient’s　tibials　anterior　(TA)　to　achieve　foot　dorsiflexion　and　assists　the　patient　in　walking.　It　is　an　effective　way　for　foot　drop　correction.　In　order　to　realize　the　activation　mode　of　tibialis　anterior　muscle　similar　to　natural　gait　under　electrical　stimulation,　this　paper　proposed　a　natural　activation　electromyography　(EMG)　prediction　algorithm　based　on　dynamic　BP　neural　network,　which　uses　angular　velocity　of　gait　to　predict　EMG　signal　in　real　time,　and　then　modulate　FES　output　intensity.　Based　on　STM32F407　microcontroller,　a　wearable　self-adaptive　EMG　modulation　functional　electrical　stimulation　therapy　device　was　developed.　Seven　healthy　people　were　recruited　for　establishment　of　EMG　prediction　model,　and　a　foot　drop　patient　was　recruited　for　an　FES　gait　experiment.　The　electrical　evoked　muscle　response　produced　by　FES　was　compared　with　the　natural　activation　of　healthy　subjects　under　natural　gait.　The　correlation　coefficient　between　the　two　reaches　0.7490 ± 0.1185,　which　is　higher　than　the　traditional　trapezoidal　profiles　stimulation　method　(0.5318 ± 0.1418).　The　results　show　that　the　control　system　is　more　capable　of　assisting　patients　to　walk　in　a　near-natural　gait.　©　2022,　The　Author(s),　under　exclusive　license　to　Springer　Nature　Singapore　Pte　Ltd.