Week 4

Welcome back everybody, I hope you had a good week. My week was very eventful as opposed to last week. With spring break coming up I am sure everyone had their fair share of assignments and tests, and that goes for me too.

Midterms, that is the one word to sum up college life right now. Every student on ASUs campus was looking forward to the break, but had to make it through the week first. I had my share of work to do too. In the spirit of midterms, I took a test in Dr. Berisha's class on Thursday. Granted it did not count for anything, I still tried to do well. The test was on Fourier series and transforms, so when I get my grade back we'll see if I was qualified to explain it to you guys last week.

Along with the test, I also got my first assignment/task for the cough project. Prad said that one of the problems with the project was the limited amount of data. With the machine learning based algorithm that they are working on, the more data the better; but at the moment they only have samples they recorded themselves (meaning fake coughs) and a 24 hr sample a doctor in Manchester gave them. Now, where do you start to look for real world samples of people coughing... YouTube right? But if you search for cough on YouTube you'll get some weird things, so I had to be more specific. That's when I remembered that Hilary Clinton coughed a lot during her campaign. So I found a good video with a montage of her coughing in speeches.

What did I do with this sample? First I had to convert it from stereo audio to mono audio, because the device only records one channel of audio. To do this I wrote a short program in MATLAB which took the audio file in mp3 format, averaged the two channels to make one, then exported it in wav format (just another way to store audio). Once this was complete, I loaded the audio file in a new program called Audacity, a free audio editing program. Here I went through the entire sample and labeled every cough with a start and end time (I'll put a picture of what I did). With those labels I can run the audio through the algorithm and either train it or test how well it has been trained, which I will do next week. I was going to try to put an audio player of the nine seconds shown in the picture, but it isn't possible, so I uploaded it on sound cloud and here's the link: nine second sample

The red circle is a label; the blue circle is an example of a cough; and the section of yellow is when there was a lot of noise (cheering) in the background.

This assignment was a pretty exciting step for me. I was able to make an actual contribution to the cough project by providing more data, and now I know what I am capable of doing. Hopefully this will branch into more and more things for me to do, and I will keep you updated on all my assignments.

On that note, I'm sure you are all eager to go into spring break, and so am I. That means that next week I will not be posting, but I will still be doing a little bit of work. So go have fun during the break, and I'll see you in a couple weeks.

Buh Bye