People can achieve rich musical expression through vocal sound -- see for example human beatboxing, which achieves a wide timbral variety through a range of extended techniques. Yet the vocal modality is under-exploited as a controller for music systems. If we can analyse a vocal performance suitably in real time, then this information could be used to create voice-based interfaces with the potential for intuitive and fulfilling levels of expressive control. Conversely, many modern techniques for music synthesis do not imply any particular interface. Should a given parameter be controlled via a MIDI keyboard, or a slider/fader, or a rotary dial? Automatic vocal analysis could provide a fruitful basis for expressive interfaces to such electronic musical instruments. The principal questions in applying vocal-based control are how to extract musically meaningful information from the voice signal in real time, and how ...

The localization and tracking of sound sources using microphone arrays is a problem that, even if it has attracted attention from the signal processing research community for decades, remains open. In recent years, deep learning models have surpassed the state-of-the-art that had been established by classic signal processing techniques, but these models still struggle with handling rooms with strong reverberations or tracking multiple sources that dynamically appear and disappear, especially when we cannot apply any criteria to classify or order them. In this thesis, we follow the ideas of the Geometric Deep Learning framework to propose new models and techniques that mean an advance of the state-of-the-art in the aforementioned scenarios. As the input of our models, we use acoustic power maps computed using the SRP-PHAT algorithm, a classic signal processing technique that allows us to estimate the acoustic energy ...

Music is built from sound, ultimately resulting from an elaborate interaction between the sound-generating properties of physical objects (i.e. music instruments) and the sound perception abilities of the human auditory system. Humans, even without any kind of formal music training, are typically able to ex- tract, almost unconsciously, a great amount of relevant information from a musical signal. Features such as the beat of a musical piece, the main melody of a complex musical ar- rangement, the sound sources and events occurring in a complex musical mixture, the song structure (e.g. verse, chorus, bridge) and the musical genre of a piece, are just some examples of the level of knowledge that a naive listener is commonly able to extract just from listening to a musical piece. In order to do so, the human auditory system uses a variety of cues ...

When listening to music, some humans can easily recognize which instruments play at what time or when a new musical segment starts, but cannot describe exactly how they do this. To automatically describe particular aspects of a music piece – be it for an academic interest in emulating human perception, or for practical applications –, we can thus not directly replicate the steps taken by a human. We can, however, exploit that humans can easily annotate examples, and optimize a generic function to reproduce these annotations. In this thesis, I explore solving different music perception tasks with deep learning, a recent branch of machine learning that optimizes functions of many stacked nonlinear operations – referred to as deep neural networks – and promises to obtain better results or require less domain knowledge than more traditional techniques. In particular, I employ ...

This thesis addresses the development of a system for pitch-informed solo and accompaniment separation capable of separating main instruments from music accompaniment regardless of the musical genre of the track, or type of music accompaniment. For the solo instrument, only pitched monophonic instruments were considered in a single-channel scenario where no panning or spatial location information is available. In the proposed method, pitch information is used as an initial stage of a sinusoidal modeling approach that attempts to estimate the spectral information of the solo instrument from a given audio mixture. Instead of estimating the solo instrument on a frame by frame basis, the proposed method gathers information of tone objects to perform separation. Tone-based processing allowed the inclusion of novel processing stages for attack re nement, transient interference reduction, common amplitude modulation (CAM) of tone objects, and for better ...

Despite a lot of progress in speech separation, enhancement, and automatic speech recognition realistic meeting recognition is still fairly unsolved. Most research on speech separation either focuses on spectral cues to address single-channel recordings or spatial cues to separate multi-channel recordings and exclusively either rely on neural networks or probabilistic graphical models. Integrating a spatial clustering approach and a deep learning approach using spectral cues in a single framework can significantly improve automatic speech recognition performance and improve generalizability given that a neural network profits from a vast amount of training data while the probabilistic counterpart adapts to the current scene. This thesis at hand, therefore, concentrates on the integration of two fairly disjoint research streams, namely single-channel deep learning-based source separation and multi-channel probabilistic model-based source separation. It provides a general framework to integrate spatial and spectral cues in ...

The production and delivery of audio for television involve many creative and technical challenges. One of them is concerned with the level balance between the foreground speech (also referred to as dialogue) and the background elements, e.g., music, sound effects, and ambient sounds. Background elements are fundamental for the narrative and for creating an engaging atmosphere, but they can mask the dialogue, which the audience wishes to follow in a comfortable way. Very different individual factors of the people in the audience clash with the creative freedom of the content creators. As a result, service providers receive regular complaints about difficulties in understanding the dialogue because of too loud background sounds. While this has been a known issue for at least three decades, works analyzing the problem and up-to-date statics were scarce before the contributions in this work. Enabling the ...

This thesis concentrates on a major problem within audio signal processing, the separation of source signals from musical mixtures when only a single mixture channel is available. Source separation is the process by which signals that correspond to distinct sources are identified in a signal mixture and extracted from it. Producing multiple entities from a single one is an extremely underdetermined task, so additional prior information can assist in setting appropriate constraints on the solution set. The approach proposed uses prior information such that: (1) it can potentially be applied successfully to a large variety of musical mixtures, and (2) it requires minimal user intervention and no prior learning/training procedures (i.e., it is an unsupervised process). This system can be useful for applications such as remixing, creative effects, restoration and for archiving musical material for internet delivery, amongst others. Here, ...

Typical acoustic scenes consist of multiple superimposed sources, where some of them represent desired signals, but often many of them are undesired sources, e.g., interferers or noise. Hence, source separation and extraction, i.e., the estimation of the desired source signals based on observed mixtures, is one of the central problems in audio signal processing. A promising class of approaches to address such problems is based on Independent Component Analysis (ICA), an unsupervised machine learning technique. These methods enjoyed a lot of attention from the research community due to the small number of assumptions that have to be made about the considered problem. Furthermore, the resulting generalization ability to unseen acoustic conditions, their mathematical rigor and the simplicity of resulting algorithms have been appreciated by many researchers working in audio signal processing. However, knowledge about the acoustic scenario is often available ...

Playing techniques are expressive elements in music performances that carry important information about music expressivity and interpretation. When displaying playing techniques in the time-frequency domain, we observe that each has a distinctive spectro-temporal pattern. Based on the patterns of regularity, we group commonly-used playing techniques into two families: pitch modulation-based techniques (PMTs) and pitch evolution-based techniques (PETs). The former are periodic modulations that elaborate on stable pitches, including vibrato, tremolo, trill, and flutter-tongue; while the latter contain monotonic pitch changes, such as acciaccatura, portamento, and glissando. In this thesis, we present a general framework based on the scattering transform for playing technique recognition. We propose two variants of the scattering transform, the adaptive scattering and the direction-invariant joint scattering. The former provides highly-compact representations that are invariant to pitch transpositions for representing PMTs. The latter captures the spectro-temporal patterns exhibited ...

Estimating object motion is one of the key components of video processing and the first step in applications which require video representation. Visual object tracking is one way of extracting this component, and it is one of the major problems in the field of computer vision. Numerous discriminative and generative machine learning approaches have been employed to solve this problem. Recently, correlation filter based (CFB) approaches have been popular due to their computational efficiency and notable performances on benchmark datasets. The ultimate goal of CFB approaches is to find a filter (i.e., template) which can produce high correlation outputs around the actual object location and low correlation outputs around the locations that are far from the object. Nevertheless, CFB visual tracking methods suffer from many challenges, such as occlusion, abrupt appearance changes, fast motion and object deformation. The main reasons ...

Recent years have witnessed a significant interest in interpretable machine learning (IML) research that develops techniques to analyse machine learning (ML) models. Understanding ML models is essential to gain trust in their predictions and to improve datasets, model architectures and training techniques. The majority of effort in IML research has been in analysing models that classify images or structured data and comparatively less work exists that analyses models for other domains. This research focuses on developing novel IML methods and on extending existing methods to understand machine listening models that analyse audio. In particular, this thesis reports the results of three studies that apply three different IML methods to analyse five singing voice detection (SVD) models that predict singing voice activity in musical audio excerpts. The first study introduces SoundLIME (SLIME), a method to generate temporal, spectral or time-frequency explanations ...

Single channel source separation is a quite recent problem of constantly growing interest in the scientific world. However, this problem is still very far to be solved, and even more, it cannot be solved in all its generality. Indeed, since this problem is highly underdetermined, the main difficulty is that a very strong knowledge about the sources is required to be able to separate them. For a grand class of existing separation methods, this knowledge is expressed by statistical source models, notably Gaussian Mixture Models (GMM), which are learned from some training examples. The subject of this work is to study the separation methods based on statistical models in general, and then to apply them to the particular problem of separating singing voice from background music in mono recordings of songs. It can be very useful to propose some satisfactory ...

Audio effects modeling is the process of emulating an audio effect unit and seeks to recreate the sound, behaviour and main perceptual features of an analog reference device. Audio effect units are analog or digital signal processing systems that transform certain characteristics of the sound source. These transformations can be linear or nonlinear, time-invariant or time-varying and with short-term and long-term memory. Most typical audio effect transformations are based on dynamics, such as compression; tone such as distortion; frequency such as equalization; and time such as artificial reverberation or modulation based audio effects. The digital simulation of these audio processors is normally done by designing mathematical models of these systems. This is often difficult because it seeks to accurately model all components within the effect unit, which usually contains mechanical elements together with nonlinear and time-varying analog electronics. Most existing ...

Single-channel source separation for radio-frequency (RF) systems is a challenging problem relevant to key applications, including wireless communications, radar, and spectrum monitoring. This thesis addresses the challenge by focusing on data-driven approaches for source separation, leveraging datasets of sample realizations when source models are not explicitly provided. To this end, deep learning techniques are employed as function approximations for source separation, with models trained using available data. Two problem abstractions are studied as benchmarks for our proposed deep-learning approaches. Through a simplified problem involving Orthogonal Frequency Division Multiplexing (OFDM), we reveal the limitations of existing deep learning solutions and suggest modifications that account for the signal modality for improved performance. Further, we study the impact of time shifts on the formulation of an optimal estimator for cyclostationary Gaussian time series, serving as a performance lower bound for evaluating data-driven methods. ...