Crossbill Heartbeat Filtering and Peak Counting

Megan Black, Violet Kurtz, Kristina Marquez, Simon Millsap

The filtering algorithm works to detect the most probable carrier frequency and extracts it. Then, the signal is demodulated to find the real heartbeat signal among the noisy data. Filtering of the data closely mimics the manipulation of the data that the original researcher performed in order to more clearly detect the peaks for hand counting. The peak counting algorithm utilizes a function in the Python library SciPy that takes in sound data along with a few data parameters, and determines the number of peaks in the data. With the number of peaks, it is a simple calculation to determine the average beats per minute of any segment of data. The project was determined to be successful and fairly accurate according to the testing performed.

Program Requirements

Hardware:

Laptop or desktop that can run Python (directions written for MacOS, please reach out if using a different operating system)

Software:

Python 3, Pip

Dependencies:

Tkinter, matplotlib, numpy, scipy.io, math, PIL (version numbers viewable in requirements.txt)

Usage

To use this program

1. Download zip from this repository: https://github.com/OniLink/Songbird.git

• Click on green “Code” button
• Download zip
• Unzip the zip file

2. Navigate to the folder of the uncompressed zip file, right click the folder and select “New Terminal at Folder” to open the command prompt (commonly referred to as a terminal) like so:
3. Into the command prompt, type chmod +x setup.sh and press enter.
4. Then type ./setup.sh and press enter. Some things will begin to install.
5. After that everything should be set up! Anytime you want to open up this application in the future, simply open up a command prompt at this location as before and just type ./setup.sh
6. To actually run the gui, in the command prompt just type python gui.py and press enter.
7. To find a .wav file on your computer to use, click the “Search” button
8. Click “Filter” to filter a file, or “Read” to view an already filtered file.
9. Adjust range to fit interval needs, then click “Adjust Range” to refresh interval
10. Adjust sliders so that it looks like all the peaks in interval are being counted (Tips on this step are stated below)
11. The program will output a file called filtered_file.wav which is available for you to use if you want to view the file again with the “Read” button, or for further inspection in tools like Audacity if desired.

Sliders

Period:

This slider adjusts the period where bird heartbeats are captured and will ignore everything in between. For adjusting, the max value of the slider makes it capture a larger range of periods at the cost of catching more false peaks.

Prominence:

This slider adjusts what the code looks for in the difference between the highest peak and the median line, this helps the code pick out the peaks with more ease. It is recommended to keep this at 1 for well filtered data, and to adjust it to lower values when the data is more noisy. The max value allows for more of a difference in peak heights to be detected, the lower the value, the more lenient the prominence will be.

Tip:

A recommendation for sifting through the data is that when the data looks irregular, it is best to use a smaller interval of time

Demo Link

https://www.youtube.com/watch?v=bSxV2Pfvxkk

Unrealized Features:

• Scrollable Graph: We had planned to add a way for users to scroll through the graph of the data, but we were unable to implement this feature.
• Accounting for Crossbill Double Peaks: Tuning the slider can account for double peaks, but if the slider is tuned for single peaks it will skip past double peaks. There is no specific implementation for double peaks as of now.