We propose a method for estimating dancing skills based on rhythmical factors extracted by video processing. The proposed method consists of motion analysis to obtain motion parameters of human body parts and data analysis to transform the obtained complex motion parameters into compact forms that represent the dance actions. We use a Kalman filter to increase robustness in the motion analysis. Data analysis of motion parameters was executed using Singular Value Decomposition. As a result of analyzing transformed data sequences, we confirmed that our proposed method can extract the rhythmical factors of dance that have correlation with subjective estimation results.