手势控制音乐播放器：基于CNN的计算机视觉交互实践

This project explores building a contactless music player controlled by gestures using convolutional neural networks (CNN) and WebSocket technology. It provides an end-to-end solution: capturing gestures via camera, recognizing them with a deep learning model, and converting to music control commands. This interaction is convenient for scenarios like cooking, exercise, or for users with hand mobility issues.

• Original Author/Maintainer: gurubaranr0x
• Source Platform: GitHub
• Project Name: PRODIGY_ML_04
• Original Link: https://github.com/gurubaranr0x/PRODIGY_ML_04
• Release Time: 2026-05-29

Contactless control is a key focus in human-computer interaction (HCI) to address scenarios where hands are busy or traditional input methods are不便.

Computer Vision & Gesture Capture

The system uses a camera to get video streams and extract static gestures. Focusing on static gestures reduces model complexity and improves accuracy/response speed.

CNN Model

A custom-trained CNN is used for recognition. Convolution layers extract local features (e.g., finger contours, palm shape), while pooling layers provide spatial invariance. It supports custom gesture categories like play/pause (single finger up), next (wave right), previous (wave left), volume up (palm up), volume down (palm down).

WebSocket Communication

WebSocket enables low-latency real-time communication between the recognition module and the music player, ensuring instant response to gestures—critical for music control.

1. Accessibility: Alternative interaction for users with temporary or permanent hand不便, replacing mouse/keyboard.
2. Multi-task Scenarios: Useful during cooking, exercise, or other hands-busy activities to control music without stopping.
3. Smart Home Integration: Can联动 with other devices (e.g., a 'mute' gesture pauses music and dims lights).

Light Condition Adaptability

Problem: Sensitivity to varying light (bright, dim, backlight). Optimizations: Data augmentation (add diverse light samples), adaptive preprocessing (adjust brightness/contrast), robust model architectures.

Background Interference

Problem: Complex backgrounds affect recognition accuracy. Solutions: Human segmentation (locate hand first), background subtraction, depth camera for 3D info.

Model Lightweight

Problem: Need to run smoothly on ordinary devices. Methods: Model pruning, quantization, knowledge distillation (maintain accuracy while reducing resource usage).

The project's architecture is scalable to:

• Smart Home Control: Gesture to switch lights, adjust AC temperature.
• Presentation Control: Gesture for slide turning, laser pen function.
• Game Interaction: Somatosensory input for games.
• Industrial Control: Non-contact operation in industrial environments where touching screens is不便.

PRODIGY_ML_04 combines deep learning, computer vision, and real-time communication to create an intuitive interaction experience. It's not just a tech demo but a practical tool雏形. As edge computing and model efficiency improve, such visual interaction schemes will be more widely applied, changing how we interact with digital devices.