Abstract
We propose the use of Temporal Convolutional Networks for audio-based beat tracking. By contrasting our convolutional approach with the current state-of-the-art recurrent approach using Bidirectional Long Short-Term Memory, we demonstrate three highly promising attributes of TCNs for music analysis, namely: i) they achieve state-of-the-art performance on a wide range of existing beat tracking datasets, ii) they are well suited to parallelisation and thus can be trained efficiently even on very large training data; and iii) they require a small number of weights. © 2019 IEEE

Abstract
In this paper we present SEED, a generative system capable of arbitrarily extending recorded environmental sounds while preserving their inherent structure. The system architecture is grounded in concepts from concatenative sound synthesis and includes three top-level modules for segmentation, analysis, and generation. An input audio signal is first temporally segmented into a collection of audio segments, which are then reduced into a dictionary of audio classes by means of an agglomerative clustering algorithm. This representation, together with a concatenation cost between audio segment boundaries, is finally used to generate sequences of audio segments with arbitrarily long duration. The system output can be varied in the generation process by the simple and yet effective parametric control over the creation of the natural, temporally coherent, and varied audio renderings of environmental sounds. Copyright: © 2016 First author et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License 3.0 Unported, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Abstract
In this paper, we address a subtopic of the broad domain of audio enhancement, namely musical audio bandwidth extension. We formulate the bandwidth extension problem using deep neural networks, where a band-limited signal is provided as input to the network, with the goal of reconstructing a full-bandwidth output. Our main contribution centers on the impact of the choice of low-pass filter when training and subsequently testing the network. For two different state-of-the-art deep architectures, ResNet and U-Net, we demonstrate that when the training and testing filters are matched, improvements in signal-to-noise ratio (SNR) of up to 7 dB can be obtained. However, when these filters differ, the improvement falls considerably and under some training conditions results in a lower SNR than the band-limited input. To circumvent this apparent overfitting to filter shape, we propose a data augmentation strategy which utilizes multiple low-pass filters during training and leads to improved generalization to unseen filtering conditions at test time.

Abstract
We explore the task of computational beat tracking for musical audio signals from the perspective of putting an end-user directly in the processing loop. Unlike existing “semi-automatic” approaches for beat tracking, where users may select from among several possible outputs to determine the one that best suits their aims, in our approach we examine how high-level user input could guide the manner in which the analysis is performed. More specifically, we focus on the perceptual difficulty of tapping the beat, which has previously been associated with the musical properties of expressive timing and slow tempo. Since musical examples with these properties have been shown to be poorly addressed even by state of the art approaches to beat tracking, we re-parameterise an existing deep learning based approach to enable it to more reliably track highly expressive music. In a small-scale listening experiment we highlight two principal trends: i) that users are able to consistently disambiguate musical examples which are easy to tap to and those which are not; and in turn ii) that users preferred the beat tracking output of an expressive-parameterised system to the default parameterisation for highly expressive musical excerpts. © 2021, Springer Nature Switzerland AG.

Abstract
We propose a multi-task learning approach for simultaneous tempo estimation and beat tracking of musical audio. The system shows state-of-the-art performance for both tasks on a wide range of data, but has another fundamental advantage: due to its multi-task nature, it is not only able to exploit the mutual information of both tasks by learning a common, shared representation, but can also improve one by learning only from the other. The multi-task learning is achieved by globally aggregating the skip connections of a beat tracking system built around temporal convolutional networks, and feeding them into a tempo classification layer. The benefit of this approach is investigated by the inclusion of training data for which tempo-only annotations are available, and which is shown to provide improvements in beat tracking accuracy.

Abstract
We introduce the Harmonix set: a collection of annotations of beats, downbeats, and functional segmentation for over 900 full tracks that covers a wide range of western popular music. Given the variety of annotated music information types in this set, and how strongly these three types of data are typically intertwined, we seek to foster research that focuses on multiple retrieval tasks at once. The dataset includes additional metadata such as MusicBrainz identifiers to support the linking of the dataset to third-party information or audio data when available. We describe the methodology employed in acquiring this set, including the annotation process and song selection. In addition, an initial data exploration of the annotations and actual dataset content is conducted. Finally, we provide a series of baselines of the Harmonix set with reference beat-trackers, downbeat estimation, and structural segmentation algorithms.