Abstract:

As the existing methods to detect crackles are of non-ideal detection effects and complex computation,a new method taking the advantage of high sensitivity of fractional Hilbert transform to the abnormal components of signals is proposed. In this method,for different fractional values,Hilbert transforms are employed to transform crackle signals into the signals with stepped phase shifts. Then,functions describing the correlation between the original lung sound signals and the transformed ones are obtained with respect to different fractional orders,which are considered as the features to be matched with standard templates. The detected signals with high matching coefficients are judged as crackles,while those with low matching coefficients are judged as non-crackles. Simulated results indicate that the proposed method is effective and the detection accuracy is up to 91.2%.

Key words: signal processing, signal detection, computer-aided diagnosis, feature extraction, fractional Hilbert transform