Han Ki-beom Basketball ClassDirector Lee Hyung-joo and Coach Kim Seong-gyeom announced that they have developed Korea’s first Android-based real-time basketball shooting posture analysis application that combines Vision AI technology and Finite State Machine (FSM) structure, and are currently preparing to launch the service. This project Hyungju Lee With the leader Seonggyeom Kim It is attracting attention because coaches jointly participate in planning and development, organically combining long-term field guidance experience, sports science, engineering analysis, and artificial intelligence technology. The app developed this time breaks away from the existing method of judging basketball shooting simply by success or failure, and analyzes the entire process of making a shot as a ‘flow of action’.
‘ShotAnalyzer’, the core analysis engine, defines the shooting operation as four states: STANDBY (preparation) – RAISING (lifting) – SET (set up) – RELEASE_AND_FINISH (release and follow-through), and applied an FSM structure that detects transitions between each stage in real time. Through this, it is possible to intuitively determine at what stage a player’s shooting breaks down or which sections remain stable. In particular, this app determines the shooting time by setting an elbow angle of 125 degrees or more as the release trigger standard. This standard is a value set based on the shooting mechanism presented in previous research and the release pattern repeatedly observed in actual map sites, and is evaluated as a core algorithm that can accurately capture shooting timing without heavy computation. This is significant in that it is an attempt to transform the ‘sensory explanation’ that relied on the leader’s experience into a data-based one. Technically, Google MediaPipe Tasks Vision is used to extract joint coordinates from smartphone camera images in real time, and smooth analysis performance of over 30 FPS is implemented in a mobile environment without the need for separate wearable sensors or special equipment. By applying mirroring correction logic assuming the use of a front camera, analysis accuracy based on right-handed shooters was improved, and stability and compatibility were also secured through optimization that took into account the 16KB memory page size environment introduced in the latest Android devices.
Coach Kim Seong-gyeom, who developed the program, explained, “Shooting is not a simple repetitive movement, but a skill that precisely combines angles, timing, and connection movements. This project is not about denying the leader’s experience, but is a process of translating that experience into a data language that players can understand.” Coach Kim Seong-gyeom said, “The most difficult part on the field is explaining ‘why this shot doesn’t work,’” and added, “This app will become another coach that complements the leader’s words with visual and numerical evidence.”
In the future, Ki-beom Han’s Basketball Class plans to apply the app step by step from youth basketball education to elite player training sites, and develop it into an AI-based shooting coaching platform that encompasses education, research, and on-site guidance by accumulating data on shooting accuracy, form stability, and learning transfer effects.
Keep reading
