The cough detection camera developed by the center can recognize where coughing happens, and visualize the locations, announced the Center for Noise and Vibration Control. The cough recognition camera developed can track and record information about the individual who coughed, their location, and the number of times of coughing on a real-time basis.
Earlier, a researcher at the Department of Mechanical Engineering, KAIST developed a deep-learning based cough recognition model to categorize coughing sound in real time. The classification model of coughing event is integrated with a sound camera that visualizes the location of individuals in public places. The model showed an accuracy of 87.4%, said the research team.
Equipment to find use for Public Places
The medical equipment will be useful for public places such as offices, schools, and restaurants during epidemics, opines the researcher who developed the model.
Meanwhile, fever and coughing are most relevant symptoms of respiratory diseases, of which, fever can be detected remotely using thermal cameras. That said, the technology is expected to be very useful to detect epidemic transmissions without any contact. Furthermore, the cough event classification model is integrated with a sound camera that imagines the cough event and points out the location in the video image.
Earlier, supervised learning was carried out with a convolutional neural network to develop a cough recognition model. The model conducts binary categorization providing a one-second sound profile input, which results in an output of either cough or something else.
To establish the accuracy of the model, various datasets collected from DEMAND, Audioset, TIMIT, and ETSI. Using the model, coughing and other sounds extracted from Audioset and the rest of datasets used as background noises for data augmentation.