Selfie WanD: Designing User-Defined Gestures for Selfie Stick Photography and Gesture Recognition Using a Smartphone’s Built-in IMU

This study explores the possibility of gestural interaction on selfie sticks. We conducted a gesture elicitation study (GES) for selfie stick photography with 20 participants to identify intuitive gestures under two conditions: one allowing both movement and touch of the selfie stick, and one allowing only gestures involving its movement. Based on these findings, we implemented a system using a smartphone's built-in IMU sensor and created a machine-learning model to recognize user-defined gestures in the movement-only condition. We collected data for ten gesture types through a user study (N = 10). The model was trained on data recorded in the standing condition and evaluated using leave-one-out cross-validation across participants, achieving a classification accuracy of 99.7%. Moreover, the model demonstrated robust performance under variations in user walking behavior, selfie stick length, and stick type, indicating the feasibility of gesture-based interaction for enhancing selfie stick photography.