In order to solve some shortcomings of traditional nonlinear correction methods in grating optical fiber sensor calibration,a segmented correction algorithm is put forward based on support vector regressions (SVR) of ant colony optimization (ACO).Because the search speed of t he traditional ant colony algorithm is not ideal in the signal processing,and th e precision of the least squares support vector machine regression algorithm is not high,this algor ithm combines the ant colony algorithm to search the optimal parameter principle of the least square support vector mac hine,and the sample space is piecewise regression according to the data distribution to reduce the running time of the algorithm.The parameters of each SVM can be optimized by ACO,and the consequence is put together to achieve the integ rated characteristic of the sensor.The fitness of the algorithm can reach 99.97%.This algorithm overcomes the local op timal solution problem,and has better global convergence effect.