Hardware + human study
Wearable layout, CAD iteration, sensor integration, protocol design, and RF acquisition.
Hi, my name is Youngseo.
I’m an Electrical and Computer Engineering student at Cornell University, interested in wearable biosensors and machine learning integration.
Featured research · Cornell ECE
Wearable Radio-Frequency Muscle Sensing
Edwin C. Kan Lab · Cornell University
Can we recognize speech without listening to sound?
I developed and evaluated a wearable radio-frequency sensing system that measures subtle facial and throat muscle motion, then turns those signals into word-level predictions.
01 / Role
One project, two connected systems.
Signal + ML
Signal conditioning, time-frequency features, classifiers, and word-level performance analysis.
02 / Why muscle
Speech begins in movement, before it reaches the microphone.
This approach observes muscular activity near the face and throat instead of depending on audible speech. It opens an alternative sensing path for quiet, private, and assistive interfaces.
Near-field sensing map
Left cheekRight cheekChinThroat03 / Sensing principle
Four positions capture different parts of a spoken gesture.
Transmit and receive coils form a near-field coherent sensing link. Small changes in muscle geometry shift the measured coupling, producing radio-frequency muscle signals over time.
RF Tx / RxNear-field couplingMulti-channel RMG
04 / Wearable hardware
A wearable system had to be engineered, not simply assembled.
Each build loop used the physical prototype to inform the next geometric and electrical decision.
01
CAD + print planning
Sensor placement and enclosure geometry were designed for repeatable contact.
02
First 3D print
Rapid prototype parts exposed fit and cable-routing constraints early.
03
Fit + sensor integration
Soft structures, probes, and electrode spacing were tested together.
04
Final wearable system
The revised wearable connected near-field sensing hardware to acquisition tools.
05 / Hardware iteration
Fit and motion artifacts shaped the final design.
The build was revised around real constraints: chin-sensor placement, cable motion, repeatable geometry, and comfortable wearable contact during spoken routines.
06 / Human study
A repeatable protocol turned motion into labeled RF data.
Participants completed repeated NATO-alphabet word routines while signals were collected from four facial and throat locations.
4sensing regions
3NATO-word routines
7words per routine
210labeled signals / participant
07 / RF acquisition
From body location to a synchronized digital trace.
The RF path routes sensing hardware through SMA connections and USRP acquisition, creating time-aligned RMG data ready for analysis in the software pipeline.
1Body locations2Tx / Rx coils3SMA4USRP5LabVIEW6RMG signals
08 / Signal processing + ML
A pipeline that makes an RF muscle gesture classifiable.
01Raw RMG signal
02Bandpass filter
03Segment
04Normalize
05STFT
06Feature vector
07Classifier
08Recognized word
09 / Results
88.09%
Final combined word-classification accuracy.
The final result reflects the combined evaluation across the study pipeline. It demonstrates a viable link between quiet muscle motion and discrete-word recognition.
10 / Research output
Poster, report, and a system built through iteration.
About
Engineering, but softer around the edges.
I like building things that make technical ideas feel visible, personal, and a little magical: circuits that behave like gardens, interfaces that breathe, and systems that make signal from noise.
ECE identity, dreamy visual systems, careful interaction.
Other projects
Engineering work across research, industry, and community.
ARIBIO
Built a Python-based clinical-trial data system and translated neuroscience research into analysis for a novel Alzheimer’s therapy.
PythonClinical dataNeuroscience
Engineering World Health
Prototyped sustainable hearing aids and a low-cost prosthetic concept alongside community partners in Guatemala.
PrototypingAccessibilityProduct design
Engineers for a Sustainable World
Developed the Cornell chapter website and a practical platform for projects, recruitment, and impact tracking.
Next.jsFirebaseCommunity
Experience & research
Activity, arranged as a working record.
Bioelectronics Researcher
Jun Chen Lab · UCLA
Researching rheologically adaptive magnetic bioelectronics for motion-robust throat-based speech decoding with NATO phonetic-word recognition.
Machine Learning & Circuit Verification Engineer Intern
LX Semicon · TV T-Con DDR Part
Designed circuit-verification test protocols and built ML evaluation pipelines to reduce manual checking for display driver IC reliability.
Hardware & Software Engineer
Edwin C. Kan Lab · Cornell University
Wearable radio-frequency muscle sensing for quiet speech recognition, spanning prototyping, RF data acquisition, and ML.
Computer Engineer & Pharma Analyst
ARIBIO
Automated clinical-trial data workflows and analyzed neurodegeneration research for therapeutic strategy.
Project Co-Lead
Engineering World Health · Cornell
Led sustainable assistive-device projects including Solar Ear hearing aids and OneSize Foot prosthetics.
Web Developer
Engineers for a Sustainable World · Cornell
Built the chapter website and digital infrastructure for student-led sustainable engineering work.
Bioelectrical Researcher
Cornell
Explored a jellyfish-inspired sweat patch for biosensing and bioelectrical data collection.
Community & communication
TEDx, Women Engineers International, STEM outreach
Organized technical communities and created approachable spaces for women and younger students in engineering.
Skills / Interests
Where the circuits bloom.
Hardware Mindset
- circuits
- embedded systems
- microcontrollers
- prototyping
Signal Fluency
- DSP
- waveforms
- sensing
- data visualization
Interface Craft
- React
- TypeScript
- motion
- polished product UI
Contact