Diagnosis of Respiratory Disease using Signal Processing and Machine Learning

Abstract

There are several poor communities in this world that lack access to treatment of respiratory diseases, due to less sufficient medical expertise and less availability of diagnostic devices. Respiratory diseases like pertussis, croup, bronchitis and asthma are becoming a major problem throughout the world. Analysis of cough sound helps in detection of respiratory diseases because cough is one of the most common symptoms among all respiratory diseases. These diseases are differentiated on the basis of spectral features of cough sound. Though, in this developing era of new technologies, there are existing systems in analyzing cough signal, but still there is a need in developing tool which analyses cough signal and capable of detecting the disease at earlier stages as well as monitoring the recovery of patients suffering. In this project, we have developed a method for automatic recognition of the respiratory diseases on the basis of the cough sound. This method uses the (DSP) Digital Signal Processing techniques to extract the spectral features. In particular, we are using Short Time Fourier Transform for feature extraction, which consists of Concentration Measure and Dominant Frequency. After that we are using Neural Networks to train our machine so that it can automatically diagnose the type of respiratory disease the person is having. After that, we can easily integrate it with an expert system which provides respiratory digital health services and which provide low-cost diagnostics to base populations, to connect patients with the physicians.