2 layers96.8 × 93.5 mm Har shield, the printed circuit board in 3D

Har shield

Q: How big is the Har shield PCB?

It measures 96.8 × 93.5 mm and has 2 copper layers, 1.6 mm thick.

Q: What parts does the Har shield use?

10 parts. The full list of its parts is on this page, with links to each part's symbol and footprint where PCBKit's library has them.

Q: Where can I download the Har shield KiCad files?

From the kkawo/har-project repository on GitHub, at the commit this page was made from: the board is pcb/har-shield/har-shield.kicad_pcb and its schematic 10.D1/har-shield.kicad_sch. PCBKit opens the board in the browser, with nothing to install.

Q: Can I use the Har shield design in my own project?

Its repository states no licence, so its author keeps all rights: ask kkawo before reusing the design.

kkawo/har-project

bash # 1. 环境准备 git clone https://github.com/kkawo/har-project.git cd har-project pip install -r requirements.txt

Open in PCBKit

Size: 96.8 × 93.5 mm
Copper layers: 2
Thickness: 1.6 mm
Parts: 10
Nets: 47
Pads: 69
Tracks, vias: 103, 13
Pours: 1
Routed: Yes
KiCad: KiCad 10.0 (file 20260206)
Created: 15 Jun 2026
Updated: 26 Jun 2026

Overview

Har shield is an open-source printed circuit board by kkawo, from the kkawo/har-project repository on GitHub. It is a 2-layer board of 96.8 × 93.5 mm with 10 parts on 47 nets, routed with 103 tracks and 13 vias.

Its parts include 4 connectors, 3 resistors, 2 capacitors and 1 diode.

README

README.md

HAR Project — 可穿戴人体活动识别系统

基于惯性/磁传感器的人体活动模式识别系统设计与实现。

福建理工大学智能科学与技术 | 《模式识别与统计学习》专周实训
孔子瑜（算法与工程）、李宝平（数据采集与固件） | 导师：郑积仕

项目概述

从 PCB 设计→传感器标定→数据采集→特征工程→分类器对比→严格评估的全链路 HAR 工程实践。

硬件: ESP32-S3 + MPU6050 (Acc+Gyro) + HMC5883L (Mag) → 9通道 @50Hz
活动: 7类（静坐/站立/步行/跑步/上楼/下楼/跌倒）
数据: 2被试, 566滑动窗口 (2.56s, 50%重叠)
特征: 292维（时域14+频域10/轴, 模长, 跨轴复合）
模型: 13种分类器 LOSO 系统对比
最佳: kNN (k=3,distance) + 多数表决 L=7 → Accuracy 59.5%

项目结构

text

har-project/
├── src/                    # 算法源码 (3,681行 Python)
│   ├── utils.py            #    常量与路径配置
│   ├── preprocess.py       # D3 信号预处理与窗口化
│   ├── features.py         # D4 294维特征提取
│   ├── feature_selection.py # D5 特征选择与降维
│   ├── classifiers_baseline.py # D6 基线分类器
│   ├── models_advanced.py  # D7 高级模型与调参
│   ├── evaluation.py       # D8 模型评估
│   └── d9_advanced_exploration.py # D9 进阶探索
├── data/                   # 数据资产
│   ├── raw/real/S01-S02/   # 真实数据 (14 CSV, 566窗口)
│   ├── windowed/           # 窗口化数据集
│   └── features/           # 特征矩阵
├── calib/                  # 传感器标定参数
├── firmware/               # ESP32 MicroPython 固件
├── pcb/                    # KiCad PCB 设计文件
├── reports/                # 报告与图表 (19+ PNG)
├── docs/                   # 项目文档
├── 10.D1-10.D9/            # 各阶段交付物
├── 项目深度说明.md          # 完整项目手册
└── HANDOVER.md             # 交接文档

核心结果

模型性能排名 (LOSO)

排名	模型	Accuracy	Macro F1
1	kNN (k=3,distance) + MV L=7	0.595	0.569
2	kNN (k=3,distance)	0.564	0.527
3	HMM Viterbi	0.583	0.550
4	LogisticRegression	0.525	0.490
5	MLP (128,64)	0.507	0.472
6	RBF_SVM	0.472	0.430
7	RandomForest	0.424	0.386

各类别 Recall

类别	Recall	主要问题
walk	1.000	—
upstairs	0.935	↔ downstairs
stand	0.674	↔ sit
run	0.565	↔ walk, upstairs
fall	0.500	↔ run
downstairs	0.446	↔ upstairs
sit	0.000	↔ stand (跨被试特征偏移)

统计检验

McNemar: kNN vs RBF_SVM p < 0.0001 (显著)
Bootstrap 95% CI: [0.512, 0.594]
聚类: KMeans+PCA ARI=0.364, NMI=0.516

一键复现

bash

# 1. 环境准备
git clone https://github.com/kkawo/har-project.git
cd har-project
pip install -r requirements.txt

# 2. 确认真实数据就绪
ls data/raw/real/S01/   # 应有 7 个 CSV
ls data/raw/real/S02/   # 应有 7 个 CSV

# 3. 全链路运行 (约5分钟)
python src/preprocess.py && \
python src/features.py && \
python src/feature_selection.py && \
python src/classifiers_baseline.py && \
python src/models_advanced.py && \
python src/evaluation.py && \
python src/d9_advanced_exploration.py

# 4. 生成答辩PPT
python src/d10_generate_ppt.py

# 5. 查看结果
ls reports/d9_*.png                    # D9 进阶探索图表
ls reports/d8_*.png                    # D8 评估图表
cat reports/d9_exploration_summary.json # 数值结果

技术栈

Python ≥ 3.10 | NumPy, SciPy, Pandas
scikit-learn ≥ 1.3 | Matplotlib, seaborn
hmmlearn ≥ 0.3.0 | TensorFlow ≥ 2.13
KiCad (PCB) | MicroPython (ESP32固件)

关键文档

文档	说明
项目深度说明	完整的项目手册，包含架构、数据流、决策、结果
HANDOVER	新会话接手检查清单
项目状态	当前进度与 Open Issues
答辩问答库	25+ 答辩常见问题与答案
最终报告	课程论文 (5000+字)
AI使用记录	AI 协作开发记录

团队

角色	成员
数据采集与固件	李宝平
算法与工程	孔子瑜

参考文献

张学工.《模式识别》（第四版）. 清华大学出版社
李航.《统计学习方法》（第二版）. 清华大学出版社
Bulling, A., et al. "A Tutorial on Human Activity Recognition Using Body-worn Inertial Sensors." ACM Computing Surveys, 2014.
Anguita, D., et al. "A Public Domain Dataset for Human Activity Recognition Using Smartphones." ESANN, 2013.

Parts

Qty	Value	Footprint	References
2	Conn_01x22	Connector_PinHeader_2.54mm:PinHeader_1x22_P2.54mm_Vertical	J1, J2
1	100nF	Capacitor_SMD:C_0805_2012Metric	C2
1	220	Resistor_SMD:R_0805_2012Metric	R3
1	4.7k	Resistor_SMD:R_0805_2012Metric	R2
1	C	Capacitor_SMD:C_0805_2012Metric	C1
1	Conn_01x05	Connector_PinHeader_2.54mm:PinHeader_1x05_P2.54mm_Vertical	J104
1	Conn_01x08	Connector_PinHeader_2.54mm:PinHeader_1x08_P2.54mm_Vertical	J103
1	LED	LED_SMD:LED_0805_2012Metric	D1
1	R	Resistor_SMD:R_0805_2012Metric	R1

Questions

How big is the Har shield PCB?

It measures 96.8 × 93.5 mm and has 2 copper layers, 1.6 mm thick.

What parts does the Har shield use?

10 parts. The full list of its parts is on this page, with links to each part's symbol and footprint where PCBKit's library has them.

Where can I download the Har shield KiCad files?

From the kkawo/har-project repository on GitHub, at the commit this page was made from: the board is pcb/har-shield/har-shield.kicad_pcb and its schematic 10.D1/har-shield.kicad_sch. PCBKit opens the board in the browser, with nothing to install.

Can I use the Har shield design in my own project?

Its repository states no licence, so its author keeps all rights: ask kkawo before reusing the design.

Project files

Show this board on your page

A live frame, for any web page

<iframe src="https://pcbkit.io/embed/boards/kkawo-har-project" title="Har shield in PCBKit" width="100%" height="480" style="border:0;border-radius:12px" allow="fullscreen" loading="lazy"></iframe>

A picture

<img src="https://pcbkit.io/embed/boards/kkawo-har-project.jpg" alt="Har shield" width="1200" height="900">

A picture that opens the board, for a README

[![Har shield](https://pcbkit.io/embed/boards/kkawo-har-project.jpg?size=s)](https://pcbkit.io/boards/kkawo-har-project)

Source

Repository: kkawo/har-project
Board file: pcb/har-shield/har-shield.kicad_pcb
Commit: eb8192d
Licence: Not stated
Copies: 2 copies in 1 repository

Similar boards

2 layers57 × 55.2 mm
LogicAnalyzer
gusmanb/logicanalyzer
24 channel, 100Msps logic analyzer hardware and software
10 parts5.1kGPL-3.0
2 layers32.2 × 15.4 mm
Protector
byrantech/laptop
anyon_e, a highly integrated, high end, open source laptop. Attempt the impossible.
10 parts4.3kMIT
2 layers14.4 × 10.1 mm
Q3Q SWAB A1-00 Temperature + Test Points + OPT3001
joeycastillo/Sensor-Watch
A board replacement for the classic Casio F-91W wristwatch
10 parts2.1k
2 layers26 × 18 mm
Conn
yaqwsx/KiKit
Automation tools for KiCAD
10 parts2.1kMIT
2 layers44.3 × 30.5 mm
Lime2-SATA RevA
OLIMEX/OLINUXINO
OLINUXINO is Open Source / Open Hardware, low cost from EUR 24 Linux Industrial grade Single Board Computer capable to operate -25+85C
10 parts1.5k
2 layers150 × 50 mm
Breakout 55pin 963063-15-Connector
rusefi/rusefi
rusefi - GPL internal combustion engine control unit
10 parts1.2k
2 layers17.8 × 43.2 mm
Nano Adapter
VoronDesign/Voron-Hardware
Cool Hardware designed by the Voron Design group
10 parts1.1kGPL-3.0
2 layers53.9 × 23.2 mm
Buffer
hoglet67/RGBtoHDMI
Bare-metal Raspberry Pi project that provides pixel-perfect sampling of Retro Computer RGB/YUV video and conversion to HDMI
10 parts1kGPL-3.0