A　series　of　electrochemical　reactions　will　occur　in　the　long-term　use　of　lithium　ion　batteries,　which　will　make　the　battery　capacity　decline,　increase　the　internal　resistance,　and　even　cause　catastrophic　consequences　such　as　explosion　and　fire.　As　an　important　part　of　battery　management　system　(BMS),　accurate　estimation　of　state　of　health　(SOH)　can　improve　the　performance,　service　life　and　safety　of　batteries.　Different　from　the　usual　cycle　times　as　input　to　estimate　SOH,　this　paper　takes　the　time　interval　of　an　equal　charging　voltage　difference　extracted　from　the　constant　current　charging　curve　as　the　inputs,　and　analyzes　the　correlation　between　the　selected　feature　and　SOH　by　using　the　gray　correlation　degree　method.　Gaussian　process　regression　algorithm　is　used　to　estimate　SOH.　Mean　Absolute　Error　(MAE)　and　Root　Mean　Squared　Error　(RMSE)　are　used　to　evaluate　the　estimation　results,　and　it　is　showed　that　this　method　can　estimate　SOH　with　satisfying　accuracy.　©　2019　IEEE.