Measurement and analysis of breath sounds

• Main Author: Qiu, Yihong
• Published: University of Glasgow 2003
• Subjects: 615.84; TJ Mechanical engineering and machinery : QP Physiology
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.493796

Existing breath sound measurement systems and possible new methods have been critically investigated. The frequency response of each part of the measurement system has been studied. Emphasis has been placed on frequency response of acoustic sensors; especially, a method to study a diaphragm type air-coupler in contact use has been proposed. Two new methods of breath sounds measurement have been studied: laser Doppler vibrometer and mobile phones. It has been shown that these two methods can find applications in breath sounds measurement, however there are some restrictions. A reliable automatic wheeze detection algorithm based on auditory modelling has been developed. That is the human’s auditory system is modelled as a bank of band pass filters, in which the bandwidths are frequency dependent. Wheezes are treated as signals additive to normal breath sounds (masker). Thus wheeze is detectable when it is above the masking threshold. This new algorithm has been validated using simulated and real data. It is superior to previous algorithms, being more reliable to detect wheezes and less prone to mistakes. Simulation of cardiorespiratory sounds and wheeze audibility tests have been developed. Simulated breath sounds can be used as a training tool, as well as an evaluation method. These simulations have shown that, under certain circumstance, there are wheezes but they are inaudible. It is postulated that this could also happen in real measurements. It has been shown that simulated sounds with predefined characteristics can be used as an objective method to evaluate automatic algorithms. Finally, the efficiency and necessity of heart sounds reduction procedures has been investigated. Based on wavelet decomposition and selective synthesis, heart sounds can be reduced with a cost of unnatural breath sounds. Heart sound reduction is shown not to be necessary if a time-frequency representation is used, as heart sounds have a fixed pattern in the time-frequency plane.