Julius O. Smith

Prateek Verma, Julius Smith

In this paper, we present a framework for contrastive learning for audio representations, in a self supervised frame work without access to any ground truth labels. The core idea in self supervised contrastive learning is to map an audio signal and its various augmented versions (representative of salient aspects of audio like pitch, timbre etc.) to a space where they are close together, and are separated from other different signals. In addition we also explore generative models based on state of the art transformer based architectures for learning latent spaces for audio signals, without access to any labels. Here, we map audio signals on a smaller scale to discrete dictionary elements and train transformers to predict the next dictionary element. We only use data as a method of supervision, bypassing the need of labels needed to act as a supervision for training the deep neural networks. We then use a linear classifier head in order to evaluate the performance of our models, for both self supervised contrastive and generative transformer based representations that are learned. Our system achieves considerable performance, compared to a fully supervised method, with access to ground truth labels to train the neural network model. These representations, with avail-ability of large scale audio data show promise in various tasks for audio understanding tasks

Julius O. Smith

A simple approach to microphone- and speaker-arrays is described in which the microphone array is regarded as a sampling grid for the acoustic field, and the corresponding speaker-array is treated as a "spatial digital to analog converter" that reconstructs the acoustic field from its spatial samples. Advantages of this approach include ease of understanding and teaching, ease of deployment, effective practical guidelines for deployment, and significant computational savings in special cases. In particular, in the far-field case (acoustic sources many wavelengths away from a linear array of speakers) it is possible to quantize source angles slightly so that no processing per speaker is required beyond pure integer delay. Smoothly moving sources are obtained using well known delay-line interpolation techniques such as linear (cross-fading) and Lagrange (polynomial) interpolation between/among speakers. We call the far-field line-array case Planewave-Based Angle Panning (PBAP), in reference to the well-known Vector-Based Amplitude Panning (VBAP) family of techniques, some of which are derived here as special cases: When speakers undersample the acoustic field, the result may be considered a form of VBAP, and VBAP is also obtained as a limiting case of polygonal PBAP arrays truncated to the polygon perimeter. Spatial samples need not be on a linear array, leading to a simple spatial audio system we call Huygens Arrays (HA). HAs are quite general for sources located behind the speaker array, which no longer needs to be linear, and the sources are no longer restricted to the far field. Multiband and hybrid arrays employing VBAP (or stereo) and subwoofer(s) are discussed, using sampling theory to inform the choices of crossover frequencies.

Prateek Verma, Julius O. Smith

There has been fascinating work on creating artistic transformations of images by Gatys. This was revolutionary in how we can in some sense alter the 'style' of an image while generally preserving its 'content'. In our work, we present a method for creating new sounds using a similar approach, treating it as a style-transfer problem, starting from a random-noise input signal and iteratively using back-propagation to optimize the sound to conform to filter-outputs from a pre-trained neural architecture of interest. For demonstration, we investigate two different tasks, resulting in bandwidth expansion/compression, and timbral transfer from singing voice to musical instruments. A feature of our method is that a single architecture can generate these different audio-style-transfer types using the same set of parameters which otherwise require different complex hand-tuned diverse signal processing pipelines.

Julius Orion Smith, Harrison Freeman Smith

We derive closed-form expressions for the poles and zeros of approximate fractional integrator/differentiator filters, which correspond to spectral roll-off filters having any desired log-log slope to a controllable degree of accuracy over any bandwidth. The filters can be described as a uniform exponential distribution of poles along the negative-real axis of the s plane, with zeros interleaving them. Arbitrary spectral slopes are obtained by sliding the array of zeros relative to the array of poles, where each array maintains periodic spacing on a log scale. The nature of the slope approximation is close to Chebyshev optimal in the interior of the pole-zero array, approaching conjectured Chebyshev optimality over all frequencies in the limit as the order approaches infinity. Practical designs can arbitrarily approach the equal-ripple approximation by enlarging the pole-zero array band beyond the desired frequency band. The spectral roll-off slope can be robustly modulated in real time by varying only the zeros controlled by one slope parameter. Software implementations are provided in matlab and Faust.