The Possibility of Personality Extraction Using Skeletal Information in Hip-Hop Dance by Human or Machine

Abstract

The same dance can give different impressions depending on the way the dancers convey their own emotions and personality through their interpretation of the dance. Beginner dancers who are teaching themselves often search for dance videos online that match their own personality in order to practice and mimic them, but it is not easy to find a dance that suits their own personality and skill level. In this work, we examined hip-hop dance to determine whether it is possible to identify one's own dance from skeleton information acquired by Kinect and whether it is possible to mechanically extract information representing the individuality of dance. Experimental results showed that rich experienced dancers could distinguish their own dances by only skeleton information, and it was also possible to distinguish from averaged skeletal information. Furthermore, we generated features from the skeletal information of dance and clarified that individual dance can be distinguished accurately by machine learning.

Automatic Personality Assessment through Movement Analysis

“…Finally, the angular velocities were calculated for each joint and the different moments of time [ 46 , 58 ]. In particular, the angular velocity was calculated for the following angles: Elbows: Angles characterized by (Right/Left) Shoulder, (Right/Left) Elbow and (Right/Left) Wrist; Shoulders: Angles characterized by Spine shoulder, (Right/Left) Shoulder, (Right/Left) Elbow; Wrists: Angles characterized by (Right/Left) Elbow, (Right/Left) Wrist and (Right/Left) Hand; Knees: Angles characterized by (Right/Left) Hip, (Right/Left) Knee and (Right/Left) Ankle; Hips: Angles characterized by Spine Base, (Right/Left) Hip and (Right/Left) Knee Forward Leaning; Lateral Leaning.…”

“…The disadvantage of this work is that it is difficult to interpret the selected features because they are Fast Fourier Transform coefficients. Additionally, with Kinect, Furuichi et al showed that a professional dancer could identify himself/herself based on the skeleton features obtained while performing a short 15-s choreography [ 46 ]. They considered the dancers’ performance to be a reflection of personality.…”

“…Similarly to the work of Batrinca et al [ 40 ], we estimated the personality of an individual from the movements made while conducting a short interview. However, these movements were automatically captured by the Kinect camera instead of being manually registered, following the same capture methodology as the works developed by Sun et al [ 45 ] and Furuichi et al [ 46 ]. Another difference with the works of Batrinca et al [ 40 , 41 ] is that, in our work, all the joints of the body were analyzed instead of only focusing on the upper body [ 40 , 41 ].…”

The AffectMove 2021 Challenge - Affect Recognition from Naturalistic Movement Data

“…The authors obtained 0.81 and 0.64 accuracies for basic and improvised salsa steps respectively. Personal dance style has also been automatically detected from optical motion capture data captured during hip hop dances with 0.99 accuracy for beginner dancers and 0.92 accuracy for more experienced dancers [30]. The results in [30] were based on 12 dancers in either group of dancers (beginner and experienced) and hold-out validation with unseen dance instances in the test set.…”