From Ubicomp Lab - University of Washington
Coughing is the number one symptom individuals report when experiencing an illness. Existing approaches to cough assessment either require a patient to self-monitor their coughs or require wearing specialized equipment. We have developed algorithms for using audio recorded from a mobile phone microphone to count the number of cough episodes an individual has and the number of coughs within each episode. This system could be used to track and monitor cough frequency for a single person or, when networked, trends across an entire population--using nothing more than an individual's existing mobile phone.
Our system uses a form of feature dimensionality reduction on the audio spectrogram in order to classify coughs and protect the privacy of recorded speech. The algorithm allows for cough sounds to be reconstructed from the features with high fideilty, but does not reproduce speech that is intelligible.
- Larson, E., TienJui, L., Liu, S., Rosenfeld, M., Patel, S.N. (2011). Accurate and Privacy Preserving Cough Sensing using Low-cost Microphone. Proceedings of the 13th International Conference on Ubiquitous Computing (UbiComp 2011), Beijing, China, Sep 17-21, 2011. [Acceptance Rate: 16.6% (50/302)]. [pdf]