This　paper　proposes　an　integral　sliding　mode　control　that　integrates　a　high-gain　observer　(HGO)　and　a　radial　basis　function　neural　network　(RBFNN)　for　a　permanent　magnet　synchronous　motor　(PMSM)　with　uncertainty.　Since　the　second-order　motion　equation　of　the　PMSM　is　used　to　improve　the　control　performance,　the　speed　derivative,　which　cannot　be　measured　directly,　is　required.　Thus,　the　HGO　is　designed　to　estimate　the　unknown　state　(speed　derivative).　In　addition,　the　RBFNN　is　designed　to　approximate　the　compounded　disturbance　including　the　lumped　disturbance　of　system　and　the　HGO　error　effect.　Unlike　previous　studies,　the　output　of　the　RBFNN　is　compensated　by　both　the　controller　and　the　HGO　to　improve　the　system　robustness　and　observer　accuracy.　The　sliding　function　and　the　HGO　error　are　both　taken　into　account　in　the　RBFNN　to　explicitly　guarantee　the　stability　of　the　whole　system.　To　demonstrate　the　superiority　of　the　proposed　method,　comparative　simulations　and　experiments　were　carried　out　in　different　cases.　©　2020,　The　Korean　Institute　of　Power　Electronics.