Identifying Bird Songs
Hi, I’m Zach Robinson. I’m a senior in Computer Science Engineering at the University of Michigan. I’m working on the Songbird Identification System. This is an ongoing project that was started in January 2017 and will continue development throughout the year. Thus far we have developed an initial classifier model using machine learning, setup a model web server for compiling data to make available to the public and begun working on our recording setup using cheap electret microphone.
The goal of this project is to create a device that can be deployed, for at least a week at a time, that will record audio from which it can automatically detect and identify nearby birds based on their songs. One potential use of this is to track the changes in bird’s migration patterns due to climate change.
In order to accomplish this, our plan is to record audio using a microprocessor, like an Arduino, and then detect and store the bird songs that it records. Then, using a Raspberry Pi, we will run code to determine the bird species and finally send all of this data to our website using a 3G connection. All of this will be powered using a battery and solar panel, so it should run for at least a week on its own, making it easy to setup and use.
Thus far we have a microphone and amplifier circuit,
a website and database with example classifier data,
and our working classifier. 
Once we have a working prototype, we can begin developing an inexpensive kit that can be widely distributed to gather data from all across the country and the world.
In the next week or so, I will create a simple bandpass filter in hardware, which filters a particular frequency range, using this we can detect when a bird is singing in this range and, using an Arduino microprocessor, we will then record this. Tomorrow I will also be attending the ARROW Technology EXPO in Livonia, MI to do research on how our device should send its data back to us from wherever we are recording from. The intent is for the device to send recordings directly to our website to be analyzed or to store them on an sd-card to be analyzed later on. I should have a working model of this by the end of the summer. Thanks for taking in interest in this project. More updates to come soon!