Tactiq — hero

2025 · Harvard · Advised by Prof. Yoel Fink

Tactiq

E-Textile · Edge AI · Physical Computing
Sensing Technology

Role: System Design, Textile Fabrication, Circuit Design, Neural Network Training, System Integration

A woven pressure-sensing textile with 85.7% real-time posture classification at ~1 ms on-device inference.

Tactiq is a high-resolution pressure-sensing textile platform developed for multiple applications, including healthcare, interactive technologies, robotics and wearables. The project was advised by Prof. Yoel Fink at MIT Material Science and Engineering Department. The system integrates conductive yarn, velostat sensing geometries, and silicone formulations into a woven textile that captures distributed pressure data across a surface.

An ESP32 microcontroller handles data acquisition and runs a 4-class posture classification neural network trained in-house, achieving 85.7% real-time accuracy, a 0.85 weighted F1 score, and approximately 1 ms on-device inference latency, making true edge deployment viable in clinical settings without cloud dependency.

85.7%

Real-time Accuracy

0.85

Weighted F1 Score

~1ms

On-device Inference

Sensing Materials

Conductive ThreadsVelostatYarn

Hardware

ESP32Custom PCB

Software

4-class Neural NetworkEdge InferenceEasyEDA Circuit SchematicsAI Engineering

Fabrication

Machine Weaving KnittingBraidingSewing Digital Embroidery

Context

Posture MonitoringPressure MappingHealthcare and WellnessRobotics Sensing

Application - The Problem


Pressure Injuries are Preventable.
Yet Cost Healthcare Billions Every Year.

2.5M

annual cases in the U.S. alone

$70k–$150k

per severe case — among the most costly hospital-acquired conditions

0

continuous monitoring once a patient leaves the clinic

$26.8B

annual U.S. healthcare burden from pressure injuries — a largely preventable cost

The existing systems are

  • Expensive: out of reach for home or long-term care settings
  • Clinic-bound: no continuous coverage in daily life
  • Reactive: injuries are detected after damage has already occurred
Pressure injury context
Tactiq system

Application - The Phygital Solutions

A continuous pressure-sensing textile system paired with a Pressure Intelligence™ engine for proactive injury prevention.

Tactiq brings pressure monitoring into everyday life, woven into the surface, invisible to the patient, always on.

01

Continuous Monitoring

Visibility beyond the clinic: pressure data captured throughout the day, not just during check-ups.

02

Real-Time Intelligence

Posture and pressure exposure analysis processed on-device at ~1 ms latency, and no cloud required.

03

Proactive Intervention

Risk alerts and reposition guidance delivered before injury develops, shifting care from reactive to preventive.

System Design

Smart Textiles Design

Flexible woven sensing architecture enabling real-time distributed pressure intelligence.


Circuit schematic

Sensing Layer

  • Woven matrix with conductive thread electrodes provides high-resolution orthogonal pressure matrix
  • A piezoresistive film (velostat) is sandwiched between 2 fabric layers and the laminate is encapsulated through sewing

Data Acquisition

  • ESP32 scans the full sensing matrix at sampling rates sufficient for real-time classification
  • 16-channel multiplexer architectures with voltage-divider readout for pin-efficient scanning of 64–70 sensing nodes

Edge AI Classification

  • 4-class posture classifier trained on collected pressure maps using Python
  • TinyML deployment achieves ~1 ms inference — suitable for real-time clinical monitoring without cloud dependency
Circuit and Electrical Engineering

Circuit and Electrical Engineering

Woven Circuit Structure

Woven Circuit Structure

Neural Textile Sensor Architecture

From a single sensor point to a thread of Fiber to Matrix

Conductive Textile Design

  • Braided conductive yarn woven into flexible pressure-sensing fabric
  • Layered piezoresistive structure converts compression into electrical signals
  • Embedded circuitry designed for wearable integration

Fabrication & Validation

  • Compared knitting, sewing, and weaving fabrication methods
  • Woven structures produced lowest noise and highest signal stability
  • Iterated from 4×4 copper prototypes to scalable 8×8 textile matrices

Scalable Sensing Platform

  • Continuous, low-cost distributed pressure sensing
  • Real-time pressure mapping with edge AI integration
  • Scalable across gloves, footwear, seating, bedding, and rehabilitation systems
Neural textile architecture overview
Conductive textile detail Fabrication method Validation testing

Pressure Intelligence™ Engine

Embedded Edge AI for Real-Time Pressure Intelligence

Neural Network

  • 64-channel textile pressure sensor array (8×8)
  • 10 Hz real-time pressure sampling
  • 1,280-feature temporal input vectors
  • Quantized edge AI inference on-device
1280Input
128Nodes
64Nodes
32Nodes
4Output

Training

  • Trained on thousands of pressure windows
  • 4-class posture classification: Neutral, Lean Left, Lean Right, Empty
  • Validated on unseen real-world pressure data

Performance

85.7%

Real-time classification accuracy

0.85

Weighted F1 score

~1ms

Edge inference latency

  • Strong left/right pressure separation demonstrated
Pressure Intelligence Engine

Making

Fabrication & Testing


Fabrication detail Fabrication detail
← Overview All work Next Project →