Abstract:
Speech is an important tool for communication. When a person speaks, the vocal cords come closer and the glottis is
partially closed. The airflow which passes through glottis is disturbed by vocal cords and speech waveform is produced. The person who
suffers from the vocal cord paralysis or vocal cord blister, his lungs are filled with fluid and airway blockage cannot generate a similar
waveform as a healthy person. In this work, we compare traditional approaches with deep learning based approaches for respiratory
disease detection to distinguish between a healthy person and the victim of pathological voice disorder. Four conventional machine
learning classifiers and a one-dimensional convolution neural network based classifier have been implemented on two benchmark
datasets ICBHI 2017 and Coswara. Our experiments show that the CNN based approach and Random Forest algorithm exhibit superior
performance over other approaches on ICBHI 2017 and Coswara datasets, respectively.
