The　MKXL-B　core　muscle　strength　training　system　is　a　set　of　medical　equipment　which　relieve　or　cure　common　low　back　pain　in　patients.　The　prescription　design　of　traditional　muscle　strength　training　medical　device　is　usually　based　on　the　patient＇s　physical　condition　and　the　expert＇s　own　experience.　Due　to　the　patient＇s　own　particularity　and　other　reasons,　the　expert＇s　prescription　may　not　be　suitable　for　the　training　of　the　patients.　The　expert　will　comprehensively　evaluate　the　patient＇s　condition　to　design　the　parameters　of　the　prescription.　This　process　is　a　waste　of　time　and　consumes　medical　resources.　In　view　of　the　above　problems,　this　paper　presents　an　optimization　algorithm　based　on　improved　particle　swarm　optimization　and　BP　neural　network.　The　error　back　propagation(BP)　algorithm　is　used　to　realize　the　non-linear　relationship　[1]among　prescription　parameters.　However,　the　BP　algorithm　has　the　disadvantage,　which　being　trapped　in　local　minimum.　In　order　to　solve　this　problem,　an　improved　particle　swarm　algorithm(PSO)　is　proposed,　which　give　the　global　optimal　core　muscle　strength　training　prescription　output　parameters,　namely　compaction　pressure　and　relaxation　pressure.　Through　simulation　and　optimization　experiments　on　the　prescription　parameters　of　different　patients,　verifying　the　accuracy　and　effectiveness　of　the　algorithm,　at　the　same　time,　the　required　accuracy　requirements　are　basically　satisfied.　©　2018　IEEE.