Unverified50% confidenceFactExact time
Google MediaPipe Hands 类框架通过机器学习模型检测手部的 21 个关键点来识别手势
1
Sources
50%
Confidence
Long-term
Relevance
8/24/2026
First Seen
Sources
Related Entities
Related Claims
Unverified手部姿态估计技术需要系统实时追踪手指的21个关键节点，包括指尖、关节和掌根78% similarUnverified针对啃咬手指/习惯性动作纠正的手部监测手环，动作识别应基于6轴或9轴IMU（加速度+陀螺仪，含磁力计更佳），单轴3轴款误报率明显偏高，日常挥手、拿东西易被误判为咬手72% similarUnverifiedMediaPipe Hands 采用两阶段级联管线：基于 BlazePalm 的手掌检测器和手部地标回归模型72% similarUnverified握持方式决定形状：趴握(手掌贴合)选大且高背的人体工学形，指尖握选小巧轻量的对称形，握爪居中选中等长度带明显背峰的鼠标70% similarUnverifiedMediaPipe 手部追踪采用两阶段检测：先用掌纹检测器定位手掌区域，再用手部地标模型精确估计各关节三维坐标69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/795771API
curl https://kongchang.com/api/v1/knowledge/claims/795771MCP
get_claim(id=795771)