Publicaties

Over the past years, deep neural networks (DNNs) have quickly grown into the state-of-the-art technologyfor various machine learning tasks such as image and speech recognition or natural language processing.However, as DNN-based applications typically require significant amounts of computation, running DNNson resource-constrained devices still constitutes a challenge, especially for real-time applications such aslow-latency audio processing. In ...

This paper presents TarsosDSP, a framework for real-time audio analysis and processing. Most libraries and frameworks offer either audio analysis and feature extraction or audio synthesis and processing. TarsosDSP is one of a only a few frameworks that offers both analysis, processing and feature extraction in real-time, a unique feature in the Java ecosystem. The framework contains practical audio processing algorithms, it can be extended ...

Linear prediction has generally failed to make a breakthrough in audio processing, as it has done in speech processing. This is mostly due to its poor modeling performance, since an audio signal is usually an ensemble of different sources. Nevertheless, linear prediction comes with a whole set of interesting features that make the idea of using it in audio processing not far fetched, e.g., the strong ability of modeling the spectral peaks that ...

The PhD explores advancements in digital signal processing (DSP) in two core areas: broadband digital communication, specifically digital subscriber line (DSL) technology, and audio and speech processing, focusing on subband system identification. These domains, though distinct in application, share a common reliance on narrowband signal processing through filter bank techniques In broadband communication, the research addresses key challenges ...

Homeostatic Activity Dependant Structural Plasticity (HADSP) is a recently introduced technique to generate network using structural plasticity. The algorithm use only homeostatic plasticity but let emerge principles of Hebbian learning. A previous study suggested that HADSP was able to generate networks that effectively leverage the inter-relationships between correlated time series but the idea was tested only on simple benchmarks. This paper ...

The research work of the present dissertation is dedicated to the application of Bayesian or Bayesian non-parametric matrix factorization methods in decomposing an audio signal to its' components. At the first phase of the work, the effectiveness of some proposed Bayesian NMF frameworks are investigated when applied to the task of word recognition of speech utterances. These Bayesian approaches are used for estimating the number of word patterns ...

An ideal audio retrieval system efficiently and robustly recognizes a short query snippet from an extensive database. However, the performance of well-known audio fingerprinting systems falls short at high signal distortion levels. This paper presents an audio retrieval system that generates noise and reverberation robust audio fingerprints using the contrastive learning framework. Using these fingerprints, the method performs a comprehensive ...

Smart sensing has been subject of significant academic and industrial interest, as these services are becoming more popular due to integrated devices shrinking in size while maintaining their computational power. When designing an integrated solution for a given application, an important choice is to select a proper set of sensor modalities along with the amount of sensors that are needed to cover the environment. In this dissertation, ...