SPEECH PROCESSING AS A CHEST DIAGNOSIS ASSIST DEVICE.

Auscultation is one of the most widely used methods in the physical examination of the chest. Voice sounds heard over the chest walls give important clues in the identification and location of several pathological situations. A parametric representation and spectra of normal and abnormal voice sounds is given by autoregressive modeling. The acoustic transfer function of the chest is estimated using a fast GLS algorithm. The behavior of an automatic classifier based on the Mahalanobis distance is presented. The classifier can reliably distinguish between normal and pathological patients. Such an automatic classification device can be implemented on a microprocessor giving the physician a reliable low cost, diagnostic assist device.