Abstract
Measurements of breathing have been used as one of the tools used in the diagnosis of sleep apnea. Respiratory inductance plethysmography, temperature sensor and thoracic impedance cardiography have mainly been used for measuring breathing information. However these methods require the sensors attached on the body of the subjects. Under a quiet environment, the breathing information can be recorded by using non-contact microphone. The sound recordings include not only breathing but also other sound. In addition, breathing sound should be recorded under low signal-to-noise ratio (SNR). Our group has developed neural network (NN) based method to detect breathing sound in the sound recordings. This method can detect low SNR breathing sound with high accuracy. However, the method needs to use multi neural networks. For this reason, we make effective use of testing process and propose new NN-based method to detect breathing sounds with higher speed. We show that the proposed method can accurately detect low SNR breathing sound with higher speed.
