The role of anatomical skeletons in animation production is increasingly important. Crafting a customized anatomical skeleton generally involves detailed manual modeling or depends on high-precision motion capture equipment, both of which can be costly to implement. This paper presents PASI (Personalized Anatomical Skeleton Inference), a novel end-to-end method for estimating anatomical skeletons from monocular inputs. PASI incorporates two modes of pose estimation algorithm to acquire motion data, eliminating the need for special motion capture devices. For bone personalization, we introduce an innovative scaling algorithm to prevent the distortion of bone shape during transformation. Besides, an optimization step is also proposed to minimize motion error, enhancing the fidelity of the inferred skeleton. The method is evaluated for its performance and processing speed, showing promising results in generating personalized skeletons.