Orchestrate -A GAN Architectural-Based Pipeline for Musical Instrument Chord Conversion

Abstract

Acoustic instruments produce sounds that are characterized by specific patterns and qualities, including harmonic content, attack, and decay, vibrato, resonance, and timbre. The creation and manipulation of instrumental sounds in various musical contexts are one of the most important features of acoustic instruments. Acoustic music is unamplified music that produces sound only by vibrating air and acoustic means, instead of through electronic or virtual instruments. Acoustic music emphasizes simplicity in its lyrics, harmonies, and melodies. The conversion of one musical instrumental chord to another musical instrumental chord is possible in acoustic instruments. In this paper, the Differentiable Digital Signal Processing technique is employed as a new approach to the realistic neural audio synthesis of musical instruments that combines the efficiency and interpretability of classical DSP elements such as filters, oscillators, reverberation, etc. The deep learning techniques are incorporated to train the model and produce harmonious music patterns. The generated music preserves the feature of the real play. The method also allows non-instrumentalists to process music. The model can be further developed to feed existing music. The preprocessed data is fed as input to obtain the desired instrumental chord or music.