Respiratory Diseases discrimination based on acoustic lung signals and neural networks

Some studies show that Chronic Respiratory Diseases (CRD) are a critical problem of health public in developing countries. Especially, diagnosis can be a challenge for the medical staff when the resources are limited. In this way, new tools can contribute to clinicians and physicians in diagnostic tasks, supporting with additional information. In this case, lung acoustic signal was acquired and processed by Mel Frequency Cepstral Coefficients (MFCC) to obtain representative parameters for Artificial Neural Network (ANN) training. Experiments are presented, using different effects of distortion coding and transmission errors for five channels. Results show that the use of ANN maintains the results for classification despite the differences between channels. At same time, classification rate drop 10% as maximum, when these channel effects were analysed, compared with no channel distortion.