All work

WPI — BASHLab research

In-ear multimodal sensing device

Medical device R&D · June 2025 to June 2026

Project

Discipline
Hardware
Timeline
June 2025 to June 2026R&D internship
Role
R&D internWorcester Polytechnic Institute (WPI)
Category
Medical device engineeringWearable health technology
Status
PrototypeCo-authored paper, WPI BASHLab

Tools

  • AutoCAD
  • C++
  • Embedded systems
  • Medical device design
  • Biosensing
  • Prototyping

Overview

At Worcester Polytechnic Institute, I worked on the design and development of a wearable medical earbud.

The device targets the region beneath the ear near the jaw, a site shown in recent research to yield rich physiological data. From there, the earbud is designed to capture heart rate, respiration, voice and motion.

I used AutoCAD for the mechanical design and C++ for the embedded code, with a light, comfortable fit as a hard constraint.

Inside the platform

LYNX, the lab's platform, pairs an in-ear and an out-ear microphone with a 9-DOF IMU in a 3D-printed shell — all off-the-shelf parts, open-source firmware, sub-millisecond synchronization between the audio and motion streams. The out-ear microphone is a noise reference: it captures the same ambient noise as the in-ear channel, and an adaptive NLMS filter subtracts it, leaving the physiological signal — breathing, speech, chewing — for downstream analysis.

LYNX system design — circuit diagram of the dual microphones and IMU wired to a Raspberry Pi, and an exploded view of the 3D-printed earbud shell showing mic and IMU placement
System design — dual mics + 9-DOF IMU in a 3D-printed shell, no custom PCB required
Diagram of signal enhancement via reference microphone — outer mic noise and raw in-ear signal combined through an adaptive NLMS filter into a filtered in-ear signal
The signal path — the outer mic is a reference channel; NLMS subtracts the shared noise
Recording and motion test results — stereo microphone RMS aligned with accelerometer, gyroscope, and magnetometer traces during head tilts, plus raw versus NLMS-filtered signal plots
Test results — IMU head-tilt events align with in-ear audio peaks in a 10-minute recording

Key contributions

  • Worked on a wearable medical earbud designed to capture heart rate, respiration, voice and motion.
  • Worked directly with a WPI professor from concept to prototype: AutoCAD for the mechanical design, C++ for the embedded programming.
  • Co-author on the lab's LYNX paper.

Internship ended June 2026