This thesis proposes a multimodal integrated approach for automatic recognition of pi ano proficiency levels, using the extraction and analysis of morphometric, kinematic, biomechanical efficiency and acoustic features. Following a structured acquisition and processing pipeline, these features are supplied to supervised classification models designed to map each performer to a three-level ordinal scale (Beginner, Intermediate, and Ad vanced). The algorithmic predictions are systematically benchmarked against a verified nominal background portfolio derived from a structural questionnaire, which quantifies the participants’ institutional seniority and practice history. An experimental protocol was designed to acquire a dataset of 68 pianists using an Intel RealSense RGB-D camera, with 64 participants effectively included in the final analysis.