Arkie

Inspiration

Coughing is one of the most common symptoms of many diseases, for instance, pharyngitis，upper respiratory infection (URI) , also, Covid-19. However, people are lack of perceiving the frequency of coughing. People often guess how often they cough in past days. They usually ignore the cough when they are asleep or concentrating on other things. Just like the footsteps per day, we want to count the daily cough times for people and we are confident it will develop new ways to grade people's health. Arkie is pronounced similar with the sound of sneezing, and that's how we name this product.

What it does

Arkie is based on YAMNet Deep Learning model to classify the sound of cough out of other background noice. It listens to people and people's surroundings, collect the cough data and upload to the database for analysis. Users can be ease to know how many and how frequent they cough on one day, and they will get a reference for health consideration.

How we built it

We firstly trained the classification model on TensorFlow. The model is trained on our homemade sample set. Then we import the model to the Python program, and using the modules such as PyAudio, Numpy to collect and process the sound data. After all, we develop the database connector to upload the predictions after filter by the threshold value to the database. Then we build visualized dashboard of the Analysis on powerbi to handle the data.

Challenges we ran into

To collect the sound samples of coughing is not easy, maybe because of the copyright issues or the quality is too low. So we decided to record our own coughing samples set. We pretend to be the worst patients of lung's disease, cough with different volume and pitch, dry or wet. As we are** almost going to be patients**

Accomplishments that we're proud of

84% of accuracy when it is the first time that we use TensorFlow building an AI program

What we learned

Collect and process the sample dataset for machine learning The use of TensorFlow framework to train and deploy the deep learning network Collaboration and Inspiring others