02 · EchoCourt 回声球场

让看不见的挥拍动作被看见、被感知——技术隐形于球场。

智能产品设计 · 交互 · 硬件 · Solo · 2025 · 导师唐林涛

网球进阶者最大的痛点，是“动作错了却说不清错在哪”。用户访谈里反复出现这样的话——“刚学会很上瘾，连打两周，动作不规范喜提网球肘”“我感觉自己拉得不对，但说不清差在哪”“以为拉拍到位了，打完看视频天塌了”。和水平相近的人对打时，人无暇顾及动作，“能打回去就是胜利”。他们需要的不是又一次练习，而是让看不见的动作偏差被看见、被即时纠正——一套完整的改进流程，而非零散提醒。

For intermediate players, the core pain is “something's wrong but I can't tell what.” Interviews surfaced it again and again — tennis elbow from bad form, a swing that “feels off” but can't be named. They don't need more reps; they need invisible motion errors made visible and corrected in the moment.

面对“初学者要激励、职业选手要数据、进阶者要低干扰智能”三类人群，我选择聚焦进阶者，并定下核心理念——技术隐形：在保留真实击球手感的前提下，把智能藏进装置里。关键取舍在硬件形态上：最初想做一把全新的智能球拍，但会让使用方式和场景过于局限；于是转向模块化组件——传感器隐形嵌入拍底盖、摄像头挂上球场围网，把“固定视觉捕捉”与“具身动作传感”分离，既减轻数据压力、又保持了产品的轻便与不改变击球习惯。

Facing three user types, I focused on the intermediate player and set one principle — invisible technology: keep the real feel of the swing, hide the intelligence inside. The key trade-off was hardware form: instead of a whole new racket (too limiting), I split it into modular parts — a sensor in the racket's end-cap and a camera clipped to the fence — separating visual capture from embodied sensing to keep it light and habit-preserving.

设计理念源自一个洞察：人的动作不是孤立的点，而是连续的、有节奏与方向的流动。我从舞蹈理论家 Rudolph Laban 的动作符号系统、以及运动摄影先驱 Gjon Mili 的轨迹影像中获得启发——身体语言能否转化为视觉、听觉、触觉，反过来辅助学习？据此搭建技术链路：动作捕捉（YOLOv8-Pose / MediaPipe 提取人体关键点）→ 轨迹生成（Python 处理时序数据）→ 正确轨迹比对（DTW 计算与标准动作的偏差，量化肘、肩、腕的关节角度）→ 多模态反馈（视觉粒子、听觉音效、触觉震动）。硬件层用 ESP32 + MPU6050 六轴传感器实测挥拍力度，超过阈值即触发差异化震动。

The concept came from one insight: movement isn't isolated points but continuous, rhythmic flow. Drawing on Rudolph Laban's movement notation and Gjon Mili's motion photography, I built a pipeline: capture (YOLOv8-Pose/MediaPipe) → trajectory (Python) → comparison (DTW, quantifying joint angles) → multimodal feedback (particles, sound, haptics), with an ESP32 + MPU6050 measuring real swing force.

成品是一套“技术隐形”的具身智能网球系统：隐形嵌入（传感器藏进拍底盖，轻触即激活）+ 挂网即用（摄像头挂上围网、无需支架、广角覆盖全场）+ 无感体验（设备自动连接，不改变击球习惯）。反馈上，系统只在关键错误或极佳表现时介入——错误时“嗡-嗡”短促双震 + 语音“再快点”，完美时“嗡——”长震 + “完美！”，避免过度打扰。对进阶练习者，App 端的粒子更多承载社交与反思价值：练习后生成粒子精彩时刻，并根据数据生成专属“粒子人格”，把“收肘”“早转髋”等改进词写进粒子爆炸里。

The result: an embodied system where tech disappears — sensor hidden in the end-cap, camera clipped to the fence, auto-connected, habit-preserving. Haptics intervene only at key errors or peak form. For advanced players the particles carry social and reflective value: a highlight reel and a data-driven “particle persona” with fixes written into the particle bursts.

这是我第一次尽可能完整地走完一遍产品设计流程——过去的课多聚焦调研或建模，这次在八周里从一个想法出发，串起调研、分析、定义、制作，还补上了“制作”方面的经验。过程中方案几经改变：用 TouchDesigner + Kinect 做的实时粒子最终没能成为产品主体，是一处遗憾。成品也还没到完全体。但这次让我第一次理解了技术哲学的基本观点，也真正体会到“从想法到初步产品”意味着什么——怀着爱和勇气去做设计。

This was my first time running a full product-design process end to end — research, analysis, definition, making — in eight weeks. The direction shifted a few times; the TouchDesigner + Kinect real-time particles didn't make it into the final product, a real regret. It's not finished, but it taught me what it means to carry an idea to a working product.

01 · 用户背景

02 · 关键决策与取舍

03 · 过程

04 · 最终呈现