Case Study: Hyfe achieves faster, longer-lasting on-device cough detection with Edge Impulse

Hyfe Translating Coughs to Actionable Insights

Hyfe, a respiratory wellness company, wanted to turn cough audio into actionable health insights while moving more of its machine learning to fully on-device processing. To do that, the team needed smaller, lighter, and more efficient models that could run reliably on resource-constrained devices like Arm Cortex-M33-class hardware.

Using Edge Impulse, Hyfe imported its massive global cough dataset, built a streamlined audio ML pipeline, and optimized cough-detection models for embedded deployment. The result was a dramatic improvement: model delivery timelines dropped from up to two years to about 2–3 months, and battery life in test devices improved from seven hours to as much as 80 hours with Edge Impulse-optimized models.

Hyfe

Joe Brew

Chief Executive Officer