While the proverb speaks of resilience, falls can be a critical health concern, especially for the elderly. This project presents a robust fall detection system that combines wearable technology and machine learning to…
System is an open-source printed circuit board by DevPro13, from the DevPro13/Fall-Detection-and-Alarming-Sys repository on GitHub, published under the MIT licence. It is a 2-layer board of 91.4 × 78.7 mm with 16 parts on 31 nets, routed with 86 tracks and 0 vias.
It is built around the ATMEGA328P-PU. Its parts include 5 resistors, 3 capacitors, 2 diodes, 2 connectors and 1 IC. It belongs with the AVR (ATmega, ATtiny) boards of this library.
"Fall seven times, stand up eight." - Japanese Proverb
While the proverb speaks of resilience, falls can be a critical health concern, especially for the elderly. This project presents a robust fall detection system that combines wearable technology and machine learning to provide timely alerts, mitigating the risks associated with delayed assistance.
This system is built upon the research detailed in the paper "Fall Detection System for Elderly People Using LSTM", Alternate Link. (Link to full paper — https://www.nepjol.info/index.php/joeis/article/view/93497 or https://www.researchgate.net/publication/404278061_Fall_Detection_System_for_Elderly_People_Using_LSTM) It leverages a custom-designed wearable device to monitor user movement and a sophisticated deep learning model to accurately distinguish between normal daily activities and dangerous falls.
The system is architecturally divided into two main sub-systems:
The operational flow of the system is as follows:
This is the core data acquisition and processing unit of the system. It is designed to be worn on the user's shoulder—a location chosen to minimize motion impedance and ensure stable, accurate readings. The device continuously captures motion data, processes it with an onboard machine learning model, and transmits an alert if a fall is detected.
Key Components:
The intelligence of the system lies in its fall detection model. Instead of simple thresholds, this project employs a Long Short-Term Memory (LSTM) neural network, a type of recurrent neural network (RNN) ideal for analyzing sequential data like human motion.
| Metric | 1-Layer LSTM Model | 2-Layer LSTM Model |
|---|---|---|
| Accuracy | 92.19% | 97.8% |
| Sensitivity (Recall) | 38.37% | 91.86% |
| Precision | 75.00% | 87.78% |
| F1-Score | 50.77% | 89.77% |
The trained model is deployed directly onto the Raspberry Pi Pico, allowing for real-time, on-device inference without reliance on a constant internet connection.
The groundstation acts as the central alerting hub. It remains in a fixed location within the user's home, listening for signals from the wearable device. Upon receiving a fall alert, it triggers both audible and visual alarms to notify nearby caregivers.
Key Components:
This custom circuit was designed to provide a stable, regulated 5V DC supply for the groundstation from an AC source.
Components:
This project is released for academic and research purposes. Hardware designs and firmware are provided as-is.
| Qty | Value | Footprint | References |
|---|---|---|---|
| 2 | 440 | Resistor_THT:R_Axial_DIN0207_L6.3mm_D2.5mm_P10.16mm_Horizontal | R2, R3 |
| 2 | C | Capacitor_THT:C_Disc_D7.0mm_W2.5mm_P5.00mm | C2, C3 |
| 1 | 10k | Resistor_THT:R_Axial_DIN0207_L6.3mm_D2.5mm_P10.16mm_Horizontal | R1 |
| 1 | 10kR | Resistor_THT:R_Axial_DIN0207_L6.3mm_D2.5mm_P10.16mm_Horizontal | R5 |
| 1 | 10u | Capacitor_THT:CP_Radial_D5.0mm_P2.00mm | C1 |
| 1 | 5.1kR | Resistor_THT:R_Axial_DIN0207_L6.3mm_D2.5mm_P10.16mm_Horizontal | R4 |
| 1 | ATmega328P-P | Package_DIP:DIP-28_W7.62mm | U1 |
| 1 | Buzzer | Buzzer_Beeper:MagneticBuzzer_ProSignal_ABT-410-RC | BZ1 |
| 1 | Conn_01x06_Female | Connector_PinHeader_2.54mm:PinHeader_1x06_P2.54mm_Vertical | J2 |
| 1 | Crystal | Resistor_THT:R_Box_L8.4mm_W2.5mm_P5.08mm | Y1 |
| 1 | GLED | LED_THT:LED_D5.0mm | D2 |
| 1 | Input | Resistor_THT:R_Axial_DIN0411_L9.9mm_D3.6mm_P5.08mm_Vertical | J1 |
| 1 | RLED | LED_THT:LED_D5.0mm | D1 |
| 1 | SW_Push | Button_Switch_THT:SW_PUSH_6mm | SW1 |
It measures 91.4 × 78.7 mm and has 2 copper layers, 1.6 mm thick.
16 parts, built around the ATMEGA328P-PU. 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.
From the DevPro13/Fall-Detection-and-Alarming-Sys repository on GitHub, at the commit this page was made from: the board is KiCAD files/system/system.kicad_pcb and its schematic KiCAD files/Hostside_sys_pcb/Hostside_sys_pcb.kicad_sch. PCBKit opens the board in the browser, with nothing to install.
Its repository publishes it under the MIT licence: you may reuse it on that licence's terms (keep the author's notice, and for a copyleft licence share your changes alike).
DevPro13/Fall-Detection-and-Alarming-Sys
While the proverb speaks of resilience, falls can be a critical health concern, especially for the elderly. This project presents a robust fall detection system that combines wearable technology and machine learning to…
zmaker/arduino_cookbook
gli sketch del nuovo progetto Arduino Cookbook e dei video di YouTube
coseyfannitutti/discipad
17-key numpad assembled with only through hole components, including usb type-c
matthewcieplak/super-sixteen
Code and schematics for the Super Sixteen Eurorack sequencer
coseyfannitutti/romeo
Staggered 40% keyboard pcb that can be assembled with only through hole components, including USB Type-C
mattybrad/drumkid
Version 4 was the first successful working design, with full code, schematics, Gerbers, etc available on GitHub.
leomil72/LM80C
A Z80-based homebrew computer withTMS9918, AY-3-8910, Z80 CTC, Z80 PIO, and Z80 SIO
