Pitch detection in musical samples has been around since an early stage of signal processing. As signal processing devices had developed, detecting notes or chords of samples had advanced as well.
In my paper my aim was to implement a program that not only detects the frequencies of the chords’ notes in a recorded sample (e.g. a guitar chord progression), but determines which chord the notes correspond to, even in the case of a longer chord progression. My solution also involves time-domain techniques, but most of the algorithm was implemented in the frequency domain. Besides detection, I also analyze statistical methods for chord classification, and after choosing the appropriate method, I implement the last part of my system, which is responsible for chord classification. In my paper I examine the methods’ weaknesses, and how I managed to overcome those problems. Finally, I present the system itself, which I tried to implement to be versatile by setting its parameters to give satisfactory results for every sample I recorded for testing. Although the program has to be developed at some points, it solves the problem of chord detection very accurately.