Anita’s Status Report for 2/11

What did you personally accomplish this week on the project? Give files or photos that demonstrate your progress. Prove to the reader that you put sufficient effort into the project over the course of the week (12+ hours).  

Most of the time spent this week was on the research behind the pitch detection algorithms. As I have no prior experience with signal processing other than my semester in 18-290, I spent many hours studying signal processing algorithms. Professor Sullivan recommended that we look into the YIN algorithm, so I watched multiple videos and read (dense) articles on this algorithm to understand the intuition behind the YIN pitch detection algorithm. There are no photos or evidence I can use to prove that I watched the videos, but the following are the sites I visited:

• The intuition behind the YIN pitch detection algorithm
• Digital Sound Processing – Stanford
• YIN, a fundamental frequency estimator for speech and music
• GitHub of YIN implementation in C++

In addition to this, I researched the risks of using a homegrown algorithm. And this is where all my biggest concerns were confirmed. It is ridiculously complex and hard to create a fast algorithm to detect pitch. Thread after thread, website after website, I delved deep into researching the optimizations and features one would have to implement to get a working real-time pitch detection algorithm. In particular, it was this StackOverflow post that was a huge reality check for me. There are just too many factors and optimizations that the best audio modules use. Especially as someone with limited experience with signal processing and debugging signal processing algorithms, these features will be extremely difficult to implement and optimize. This reinforces the research I did on previous projects that used their own signal processing algorithms. Previous projects found it infeasible to do real-time pitch detection and fell back to post-processing. 

My group and I had a long discussion about the merits of using our homegrown algorithm versus a module. We talked about our end goals and what would be the best way to get to those end goals. I asked our advisors, Ankita and Professor Sullivan, their thoughts, and as of writing this post, I am waiting for their feedback.

In addition to researching these algorithms, we had many group discussions simply breaking this project down into its details. Flushing out the project beforehand in our project proposal presentation last week helped immensely with our discussions. 

Is your progress on schedule or behind? If you are behind, what actions will be taken to catch up to the project schedule?  

As per our Gantt chart, we are on time and everything is going as expected.

What deliverables do you hope to complete in the next week?

I hope to get a bare bones pitch detection algorithm working. It doesn’t have to be real time. Of course, this is dependent on the Professor’s feedback that we are currently waiting on.