The invention provides a method for continuous blood pressure measurement based on multi-parameter combination. Besides blood pressure calibration based on pulse wave transmission time, the method also comprehensively considers a maximum value a01 of a first-order derivative of a blood oxygen volume wave which stays in the same cycle with an electrocardio-signals, a time difference a2 between two systole peak points in two adjacent periods, a time different a3 of two minimum points, a time different a4 of two diastole peak points, a difference a5 of two dicrotic notch points, an amplitude a6 of systole peak points, an amplitude a7 of the minimum point, an amplitude a8 of diastole peak points, an amplitude a9 of dicrotic notch pints, a systole area a10, a diastole area a11, an area a12 of the blood oxygen volume waves, an area ratio a13, a difference a14 of two peak points in the same cycle, a difference a15 between the systole peak point and the minimum point in the same cycle, rise time a16, a time increment a17, an increment coefficient a18, and a reflection coefficient a19. A blood pressure model can be established by a BP neural network according to the 18 parameters. A blood pressure can be predicted according to the model.