Real-time pitch detection using resource constrained IoT device

A trained ear for the recognition of notes, intervals and chords in music is a difficult skill for beginning musicians to develop due to lack of reference of these musical aspects and boring nature of its training. A system or program could rectify this issue and allow users to train their ears for this purpose. Multiple existing algorithms can distinguish pitch and chords from given audio files, using signal analysis or neural networks. However, these solutions provide only the recognition of one of the three relevant pitch based aspects. This research contributes to the development of a system that does all three by using models based on convolutional neural networks. New, custom, datasets were created to facilitate training and evaluation of the models as well. The final models were optimized and vary in terms of accuracy but show promise to be further developed into a reliable system.