Aiming at the problem that the pedestrian navigation algorithm based on quadratic curve fitting zero-speed correction technology has low utilization rate of data samples and poor correction performance and instantaneous accuracy which cannot be optimized, an intelligent adaptive pedestrian navigation algorithm based on support vector machine is proposed, which the data of the sensors is obtained by using the optimized wavelet threshold denoising algorithm; the model trained by the SVR(support vector machine regression) is used to fit the three-dimensional velocity and the error fitting result is used as the system observations; then the intelligent estimator is formed by the SVM and the Kalman filter to estimate the system errors, thereby improving the system accuracy and reliability. The experimental verification by self-developed IMU proves that the method can accurately fit the three-dimensional velocity errors, estimate the systematic error optimally, and effectively correct the navigation data. The positioning accuracy is improved by 10.9% in complex environment. The algorithm has theoretical and engineering significance.