Abstract:
Building a quick, efficient and cheap cough classification tool is a very challenging
task. It requires a well-thought and rigorous approach. One of the main take-outs
behind this tool would be a carefully collected dataset which can serve as a golden
high-quality baseline for building an accurate machine learning model.
This study intends to build a new data collection tool and analyze the collected
dataset in order to be able to differentiate between various types of cough audio
signals. There are very limited efforts that aimed to create a high reliability cough
audio signals dataset that is quality controlled by the healthcare professionals. The
current dataset contains eight types of coughs plus an additional ’other’ type. The
primary goal is to design and build a new data collection approach for the Cough
Analyzer application and secondly to conduct a data analysis of the collected dataset
which contains 1100+ entries to date submitted by the people from around the world.