BCough: Bone-Conduction-Embedded AI Platform for On-Device Cough Detection (www.mdpi.com)

BCough, an innovative AI platform, introduces bone-conduction technology for on-device cough detection, promising a significant advancement in monitoring respiratory health. Traditional cough detection methods often rely on subjective self-reporting and bulky equipment, posing challenges for real-world applications. BCough's integration of bone-conduction microphones and inertial measurement units (IMUs) allows for accurate detection of coughs while preserving user privacy by minimizing ambient noise capture. This system operates on the MAX78000 microcontroller, utilizing a lightweight convolutional neural network (CNN) that achieves impressive metrics: 0.80 Macro-F1 and 0.89 AUC, with low power consumption of approximately 20 μJ per inference. The significance of BCough lies in its ability to provide continuous, accurate cough monitoring in everyday environments without the drawbacks of cloud dependencies or extensive privacy concerns. It distinguishes itself from prior devices by leveraging bone-conduction sensing, which is less susceptible to environmental interference and captures minimal identifiable acoustic information. With its custom hardware prototype and comprehensive multimodal dataset, BCough sets a precedent for future developments in low-power, embedded AI solutions in healthcare, particularly for chronic respiratory conditions. This approach represents a promising step towards more accessible and user-friendly health monitoring technologies.