Blind Source Separation in Polyphonic Music Recordings Using Deep Neural Networks Trained via Policy Gradients

TL;DR

This paper introduces a deep neural network-based method for blind source separation of polyphonic music, utilizing policy gradients for training and a parametric harmonic model to achieve high-quality separation without ground truth data.

Contribution

It presents a novel neural network approach trained via policy gradients for blind source separation, incorporating a parametric harmonic model and direct waveform prediction.

Findings

Achieves high-quality separation with low interference.

Effective on both acoustic and synthetic audio samples.

No preprocessing of input spectra required.

Abstract

We propose a method for the blind separation of sounds of musical instruments in audio signals. We describe the individual tones via a parametric model, training a dictionary to capture the relative amplitudes of the harmonics. The model parameters are predicted via a U-Net, which is a type of deep neural network. The network is trained without ground truth information, based on the difference between the model prediction and the individual time frames of the short-time Fourier transform. Since some of the model parameters do not yield a useful backpropagation gradient, we model them stochastically and employ the policy gradient instead. To provide phase information and account for inaccuracies in the dictionary-based representation, we also let the network output a direct prediction, which we then use to resynthesize the audio signals for the individual instruments. Due to the flexibility of the neural network, inharmonicity can be incorporated seamlessly and no preprocessing of the input spectra is required. Our algorithm yields high-quality separation results with particularly low interference on a variety of different audio samples, both acoustic and synthetic, provided that the sample contains enough data for the training and that the spectral characteristics of the musical instruments are sufficiently stable to be approximated by the dictionary.