The COVID-19 pandemic has brought profound damage to the world and mankind. Despite our tremendous efforts and dedication, the end of the pandemic is still far away. We have indeed witnessed the advancement of medical and non-medical solutions to fight the pandemic. However, they are either expensive or inaccurate, and we still lack measures to outpace the coronavirus. In this study, we propose to monitor respiratory symptoms such as coughing for early detection and timely intervention to lock up potential negative effects. We minimize the cost by using popular and non-intrusive wearable device, with a variety of sensors embedded to directly measure people’s vital signs. The measured data is inherently multimodal, which can provide rich information for data analysis and insight extraction. We demonstrate a use case for cough detection based on multimodal data with accelerometer amplitude and microphone audio. A specialized two-stage algorithm is proposed. Preliminary experiments show that the algorithm is accurate in detecting cough, does not raise false alarms, and has the least communication and computational overhead. Despite being driven by the recent pandemic, our system is designed as a universal system for respiratory symptoms, and can potentially help us overcome future public health crises.