The least square method is often used to fit the input-output curve of the surface acoustic wave（SAW）micro-force sensor

however

the global search cannot be performed

and the local optimal solution is easy to obtain. Aiming at the shortcomings of traditional methods

based on the measured data of SAW micro force sensors

the generalized regression neural networks（GRNN） are used to fit the curve. The SAW micro force sensor with lithium niobate as the piezoelectric substrate is selected

and the micro force is applied to it. The output frequency data is measured by the network analyzer. The frequency-pressure data is fitted by MATLAB R2016 b according to the basic principle of GRNN and least squares method. The simulation results show that compared with the least squares method

the error of GRNN is significantly reduced

about an order of magnitude

which can effectively improve the fitting precision.