JamBot: Music Theory Aware Chord Based Generation of Polyphonic Music\n with LSTMs

2017/11/21 by Gino Brunner, Yuyi Wang, Brunner, Gino +5
Computer Science · Neuroscience · #Music and Audio Processing #Music Technology and Sound Studies #Neuroscience and Music Perception

paper · pdf · doi:10.48550/arxiv.1711.07682

Abstract

We propose a novel approach for the generation of polyphonic music based on\nLSTMs. We generate music in two steps. First, a chord LSTM predicts a chord\nprogression based on a chord embedding. A second LSTM then generates polyphonic\nmusic from the predicted chord progression. The generated music sounds pleasing\nand harmonic, with only few dissonant notes. It has clear long-term structure\nthat is similar to what a musician would play during a jam session. We show\nthat our approach is sensible from a music theory perspective by evaluating the\nlearned chord embeddings. Surprisingly, our simple model managed to extract the\ncircle of fifths, an important tool in music theory, from the dataset.\n

Related

• Generating Music using an LSTM Network
• Bach2Bach: Generating Music Using A Deep Reinforcement Learning Approach
• Polyphonic Music Composition with LSTM Neural Networks and Reinforcement Learning
• Differential Music: Automated Music Generation Using LSTM Networks with Representation Based on Melodic and Harmonic Intervals
• Polyphonic Music Generation with Sequence Generative Adversarial Networks