MULTI-BAND MEMS RESONANT MICROPHONE ARRAY FOR CONTINUOUS LUNG-SOUND MONITORING AND CLASSIFICATION

This paper reports a piezoelectric MEMS resonant microphone array for wheezing detection in lung sounds. The microphone array is composed of resonant cantilevers with serpentine support beams, and each of the resonant microphones has an unamplified sensitivity between 34.6 and 131.4 mV/Pa at the resonant frequencies, over a frequency range of 150 - 700 Hz (where wheezing spectral density is most prominent). With the high sensitivity and signal-to-noise ratio, wheezing is distinguished in both frequency domain and time domain, and as high as 97.73% identification accuracy is achieved on automatic classification through deep machine learning, when the resonant microphone array is used to record lung sounds.