Cough detection & classification: preliminary validation of Hyfe cough detection & classification


We use multiple approaches to evaluate the accuracy of Hyfe’s cough detection algorithms and the performance of its apps in various real-world scenarios. In a clinical context, Hyfe has optimized its apps for high performance in quiet indoor settings, such as a bedroom or hospital room, for any user, and for all...

Part of Hyfe's continous research efforts, this paper presents findings from tests conducted to validate the accuracy of the Hyfe cough detection technology. The study focused on two key aspects:

  1. Live Trials: The paper describes live trials with 10 volunteers in a quiet indoor setting. Each volunteer contributed 30 coughs, and the Hyfe smartphone apps captured 97% of these coughs. Sensitivity across all smartphones tested was above 93%, with a maximum of 98%. Notably, there were no false positives, indicating perfect specificity.
  2. Playback Trials: The study also involved an experimental approach: playback trials using pre-recorded coughs. These trials achieved a high sensitivity of 91%, showing the effectiveness of the system in recognizing cough sounds.