Miniaturized IMU (inertial measurement units) arewidely integrated in wearable devices, promoting the versatileand low cost pedestrian inertial navigation technology, especiallyfor indoor environment. In recent years, AI (Artificial Intelligence)is applied to improve the performance of this technology.AI methods work with data samples, thus it is important to selecta suitable process for segmenting the inertial data sequences. Thissurvey classifies AI methods for pedestrian inertial navigationinto two categories, namely human gait driven methods andsampling frequency driven methods, according to their datasegmentation process. Human gait driven methods segment theinertial measurement sequence by gait (step or stride) eventsand learn to infer a gait vector (step/stride length and direction)given a gait segment. Sampling frequency driven methods learnto infer the user's velocity or change in position given a fixedlengthsegment of inertial measurements. The survey studies theunderlying assumptions and their validity of the two categoriesof AI methods. Two methods (SELDA and RoNIN), each froma category, are chosen for evaluation and comparison, on threetesting tracks totaling 770m, covering indoor and outdoor environment,including stairs. The experiments highlight the twomethods' advantages and limitations, supporting the theoreticalanalyses. The selected methods achieve 7m and 12m positioningerrors, respectively.