Towards a non-intrusive self-management system for asthma control using smartphones

Abstract

A noise-robust algorithm for segmentation of breath events during continuous speech is presented. The built-in microphone of a smartphone is used to capture the speech signal (voiced and breath frames) under conditions of a relatively noisy background. A template matching approach, using mel-cepstrograms, is adopted for constructing several similarity measurements to distinguish between breath and non-breath frames. Breath events will be used for lung function regression.