Neural Percussive Synthesis Parameterised by High-Level Timbral Features

Abstract

We present a deep neural network-based methodology for synthesising percussive sounds with control over high-level timbral characteristics of the sounds. This approach allows for intuitive control of a synthesizer, enabling the user to shape sounds without extensive knowledge of signal processing. We use a feedforward convolutional neural networkbased architecture, which is able to map input parameters to the corresponding waveform. We propose two datasets to evaluate our approach on both a restrictive context, and in one covering a broader spectrum of sounds. The timbral features used as parameters are taken from recent literature in signal processing. We also use these features for evaluation and validation of the presented model, to ensure that changing the input parameters produces a congruent waveform with the desired characteristics. Finally, we evaluate the quality of the output sound using a subjective listening test. We provide sound examples and the system's source code for reproducibility.

Loopnet: Musical Loop Synthesis Conditioned on Intuitive Musical Parameters

“…The Wave-U-Net architecture works with fixed-length input and output, requiring all loops to have the same length. To this end, we use the rubberband library 6 to time-stretch all the loops to 130BPM. We verified that this time-stretch did not create artifacts and that the loops were still sounding coherent after this step.…”

“…This feature represents the energy distribution across the 12-note chromatic scale commonly used in western music and is also intuitive for music makers. To model the abstract non-harmonic texture of the loop, we used the perceptually pertinent timbral features proposed by Pearce et al [6,23]. These are hardness, depth, brightness, roughness, boominess, warmth and sharpness 9 .…”

“…In parallel, recent advancements in generative deep learning based methodologies have sparked a huge interest in music generation. Research has focused on both the symbolic [1,2] and the audio domain [3][4][5][6][7][8][9]. However, except for singing voice synthesizers, where the parameters of synthesis can be clearly defined in terms This work is partially supported by the Towards Richer Online Music Public-domain Archives (TROMPA) project and partly by the European Union's Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No765068, MIP-Frontiers.…”

PocketVAE: A Two-step Model for Groove Generation and Control

“…Hellmer et al [15] introduced methods to quantify the measures of microtiming of a drum track, which is also adopted in this work. Although not performance-related, there exist other works on drums, such as beat generation [16,17], drum sound synthesis [18,19] and interface design [20,21].…”

BassNet: A Variational Gated Autoencoder for Conditional Generation of Bass Guitar Tracks with Learned Interactive Control

“…At generation time, this conditional signal can be varied as a means of user control over the generated output. Along this line, models for audio synthesis have been conditioned on pitch information [36,38], categorical semantic tags [42], lyrics [5], or on perceptual features [43,44]. Other approaches use conditioning in symbolic music generation [45], or constraints during generation instead of conditional information during training [7,46].…”