Abstract:

 Crackle signals in respiratory sounds are of strong randomness and high variability and they contain lotsof disease information.In order to detect crackle signals,the S transform is introduced and a detection algorithmbased on S transform is proposed.In this algorithm,time- frequency features of crackle signals are extracted fromthe S- transform time- frequency spectrum of respiratory sound signals. Then,after a dimension reduction,localpeaks are picked out to detect crackles. Experimental results show that the proposed algorithm is effective becauseit is of an accuracy up to 93.70%,which is higher than that of some other detection algorithms.

Key words:  signal processing, signal detection, crackle signal, computer- aided diagnosis, feature extraction, time- frequency analysis