Detecting Key Signature from Digital Audio

This project, referred to as KeySeeker, seeks to analyze a digital audio piece for its musical key. A question as simple as “what's the key of this track?” remains elementary to any trained musician; yet to develop an algorithmic solution involves both broad and deep knowledge in various domain areas. The methodology combines file processing, digital signal processing, and music theory to decode an audio file, perform fast Fourier analysis, and extract its principle notes. The system then determines the most probable key by employing a hybrid between an original scale-based model and the Krumhansl-Schmuckler model, which matches notes to a predetermined note profile. Using this approach, KeySeeker can detect the correct key with 91% accuracy within 5.8 seconds, providing a solid foundation for further developments