抄録
In this paper, we propose an automatic diagnosis system for abnormal respiratory sound using a divided pulmonary sound waveform. Our method is based on a principal component analysis-linear discriminant analysis (PCA-LDA) discriminator using spectrogram features of an expiration waveform. The discrimination accuracy of the system was examined in simulations. The sensitivity and specificity of healthy against pathological discrimination and the symptom discrimination accuracy were improved compared with a mel-frequency cepstrum coefficients-Gaussian mixture model (MFCC-GMM)-based method. Computational efficiency was also improved.
