Kinect SVM is a computer vision application that allows users to trigger operating system tasks (like launching apps or controlling volume) by drawing specific shapes on a piece of paper in front of a Kinect camera.
Technical Implementation#
1. Recognition Pipeline#
- Depth Segmentation: Utilized the Kinect’s depth sensor to isolate the user’s hand and the paper from the background scene.
- Feature Extraction: Processed the image frames to extract the contours of the hand-drawn shapes.
- Classification: Fed these features into a trained Support Vector Machine (SVM) to identify the shape (e.g., Triangle, Circle, Star).
2. Performance Optimization#
- Multithreading: Implemented a heavy multi-threaded architecture in C++. One thread handled the high-bandwidth 1080p video ingestion, while separate worker threads processed the SVM prediction logic.
- Frame Rate: Successfully maintained a stable 30 Frames Per Second (FPS) detection rate, ensuring the interaction felt “real-time” to the user.
3. Training Utility#
- Custom Tooling: Built a companion desktop application that allowed us to rapidly record new gestures and train the SVM model, making the system extensible.
Tech Stack:
- Language: C++
- Hardware: Microsoft Kinect v2
- Libraries: OpenCV (Computer Vision), LIBSVM
- OS: Windows
