MIT Senior Capstone Product Design Project
Strum was created in Product Engineering Process, MIT Mechanical Engineering Capstone class. My team of 17 MIT seniors went through the design process from brainstorming to a functioning alpha prototype with in a semester given $7000 budget. Strum is a guitar attachment that allows songwriters to transcribe directly to tab. Strum is composed of three components, a custom piezoelectric sensor saddle, a microcontroller unit and a mobile app.
Felipe Lozano-Landinez, Fernando Yordan, Garrett Schulte, Grace Li, Heather Soukup, Helen Nie, Moti Ogunsina, Perry Ha, Sherry He, Brandon Corts, Brian Yue, Chuck Xia, Garrett Parrish, Jialin Shi, Manuel Castro, Sam Matthews, Sam Matthews
To detect the notes strummed, we used piezoelectric sensor. Our custom saddle designed with 6 individual sensors one for each string to minimize crosstalk. The piezo sensors are able to pick up the mechanical vibration of the strummed strings. From the signal, we can determine frequency of the vibration and therefore the note played.
The microcontroller receives and analyzes the signal to determine the string and fret number that is played. It then sends this information to the app via Bluetooth. The highest note on standard tuning guitar is around 1200Hz. To prevent aliasing, our sampling frequency needs to be above the nyquist frequency. The pre-amp circuit added a DC bias and amplified the piezo signal so we get a clear and full input.
Software team are faced with several challenges. First we need to be able to pick up the fundamental frequency, not the overtone. Second, we need to be able to accurately detect strummed note, not acoustic resonance. With the help from the amazing Professor Derek Rowell and lab instructor Steve Banzaert, we explored three potential digital signal processing algorithms for musical pitch estimation: discrete/fast fourier transfrom, YIN autocorrelation method and phase-locked loop. We ultimately decided on YIN algorithm, due to its high accuracy and speed. We combined strum detection with frequency analysis for our alpha prototype. The implemented YIN algorithm takes less than 5ms to run. World record for the fastest guitar riff is 27 notes per second. Our algorithm is able to capture both chords and riff.