Abstract:
The state-of-the-art approaches from the musical information retrieval offer many options to develop a system towards chord recognition. Audio signals are known to be featurelessly raw to derive any musical chord analysis. In addition to such, an analysis result of digital signal processing may not produce sound features that deterministically deduce any musical chords. Consequently, this project planned to create a real-time software application of chord detection using a novel methodology in machine learning, the neural networks, and a feasible approach in digital signal processing, the spectral analysis.
With a sequence of spectral analysis system, the result of such a system acts as a neural input that determines the figurable neural network model towards the objective of this project. This project was achieved to handle the chord detection of audio signals originating from real-time sources or audio files. Furthermore, this project unraveled a neural model with a desirable outcome and a sensibly less error. The system incorporates an interactive interface that provides the functionalities while acting as a bridge for the embedded chord detection.
