An identity recognition method includes a step of storing multiple training data sets that respectively correspond to gaits of multiple members, a step of obtaining a gait similarity estimation model that is established by training a Siamese neural network using the training data sets, a step of obtaining a piece of inertial data that corresponds to a step taken by a user, a step of generating, based on the piece of inertial data, a piece of feature data that is related to gait of the step taken by the user, a step of generating, based on the piece of feature data and the training data sets, multiple evaluation results that respectively correspond to the training data sets by using the gait similarity estimation model, and a step of determining whether the user belongs to a group consisting of the members based on the evaluation results and a predefined threshold. |