Real-time Guitar Chord Recognition Using Keypoint Detection and Neural Network-Based Inter-Keypoint Distance Classifier

This research addresses the significant challenge of real-time guitar chord recognition, focusing on improving accuracy in finger placement for chord playing. Real-time guitar chord recognition poses a significant challenge, particularly in accurately detecting and classifying chords during dynamic playing sessions. Existing methods often struggle with variations in finger placement, chord shapes, and orientations, leading to lower accuracy and usability in practical scenarios. This research aims to develop a robust and efficient chord recognition model capable of operating seamlessly in real-time, enhancing the overall guitar playing experience for musicians. The proposed solution combines keypoint detection and a neural network-based model, with the usage of inter-keypoint distances achieving an impressive 88.24% accuracy for a 7-chord classifier in testing. Google's Mediapipe framework for keypoint detection is utilized to address the complexities of real-time chord recognition. The study evaluates the model's performance in handling non-horizontal guitar orientations and distinguishing chords with similar shapes, demonstrating significant advancements in real-time chord recognition accuracy and usability. By addressing the complexities of real-time chord recognition, this research contributes to enhancing the usability and accuracy of chord recognition systems in practical guitar playing scenarios, benefiting musicians and music enthusiasts alike. Further exploration and refinement of this approach are recommended for continued improvement in real-time guitar chord recognition technology.