Frequency-based Skill Analysis for Motion Pictures

摘要

This paper presents sports skill classification using time series motion pictures, focused on volleyball. Our research focuses on volleyball attack, and analyzes motion picture data. We record the data using hi-speed cam-coder. We use picture pixel data but physical data as body or skeleton models. The motion picture data are transferred to the time series data with marking points, and analyzed by FFT (Fast Fourier Transform) and MFCCs (Mel-frequency cepstral coefficients) as pre-process of data mining methods as clustering and logistic regression. We attempt to classify technical skills of volleyball attacks and discuss analysis results.