Damián Martínez-Muñoz

Transient modeling by matching pursuits with a wavelet dictionary for parametric audio coding

In this letter, we propose a novel matching pursuit-based method for transient modeling with application to parametric audio coding. The overcomplete dictionary for the matching pursuit is composed of wavelet functions that implement a wavelet-packet filter bank. The proposed transient modeling method is suitable to be integrated into a parametric audio coder based on the three-part model of sines, transients, and noise (STN model). Comparative analysis between wavelet and exponentially damped sinusoidal functions are shown in experimental results. The mean-squared-error performance of the proposed approach is better than that obtained with damped sinusoids.

Tempo Driven Audio-to-Score Alignment Using Spectral Decomposition and Online Dynamic Time Warping

In this article, we present an online score following framework designed to deal with automatic accompaniment. The proposed framework is based on spectral factorization and online Dynamic Time Warping (DTW) and has two separated stages: preprocessing and alignment. In the first one, we convert the score into a reference audio signal using a MIDI synthesizer software and we analyze the provided information in order to obtain the spectral patterns (i.e., basis functions) associated to each score unit. In this work, a score unit represents the occurrence of concurrent or isolated notes in the score. These spectral patterns are learned from the synthetic MIDI signal using a method based on Non-negative Matrix Factorization (NMF) with Beta-divergence, where the gains are initialized as the ground-truth transcription inferred from the MIDI. On the second stage, a non-iterative signal decomposition method with fixed spectral patterns per score unit is used over the magnitude spectrogram of the input signal resulting in a distortion matrix that can be interpreted as the cost of the matching for each score unit at each frame. Finally, the relation between the performance and the musical score times is obtained using a strategy based on online DTW, where the optimal path is biased by the speed of interpretation. Our system has been evaluated and compared to other systems, yielding reliable results and performance.

Constrained non-negative matrix factorization for score-informed piano music restoration

In this work, we propose a constrained non-negative matrix factorization method for the audio restoration of piano music using information from the score. In the first stage (instrument training), spectral patterns for the target source (piano) are learned from a dataset of isolated piano notes. The model for the piano is constrained to be harmonic because, in this way, each pattern can define a single pitch. In the second stage (noise training), spectral patterns for the undesired source (noise) are learned from the most common types of vinyl noises. To obtain a representative model for the vinyl noise, a cross-correlation-based constraint that minimizes the cross-talk between different noise components is used. In the final stage (separation), we use the trained instrument and noise models in an NMF framework to extract the clean audio signal from undesired non-stationary noise. To improve the separation results, we propose a novel score-based constraint to avoid activations of notes or combinations that are not present in the original score. The proposed approach has been evaluated and compared with commercial audio restoration softwares, obtaining competitive results.

Voicing detection based on adaptive aperiodicity thresholding for speech enhancement in non-stationary noise

In this study, the authors present a novel voicing detection algorithm which employs the well-known aperiodicity measure to detect voiced speech in signals contaminated with non-stationary noise. The method computes a signal-adaptive decision threshold which takes into account the current noise level, enabling voicing detection by direct comparison with the extracted aperiodicity. This adaptive threshold is updated at each frame by making a simple estimate of the current noise power, and thus is adapted to fluctuating noise conditions. Once the aperiodicity is computed, the method only requires a small number of operations, and enables its implementation in challenging devices (such as hearing aids) if an efficient approximation of the difference function is employed to extract the aperiodicity. Evaluation over a database of speech sentences degraded by several types of noise reveals that the proposed voicing classifier is robust against different noises and signal-to-noise ratios. In addition, to evaluate the applicability of the method for speech enhancement, a simple F 0-based speech enhancement algorithm integrating the proposed classifier is implemented. The system is shown to achieve competitive results, in terms of objective measures, when compared with other well-known speech enhancement approaches.

Compositional model for speech denoising based on source/filter speech representation and smoothness/sparseness noise constraints

Semi-supervised NMF with source/filter speech model and constrained noise parameters.Evaluated on the 3rd CHiME and SiSEC 2013 datasets for speech and noise separation.Proposed noise constraints help to improve the isolation of speech with real noises.All tested environments but one could be modeled by the constraints.Better separation results than conventional semi-supervised sparse NMF. We present a speech denoising algorithm based on a regularized non-negative matrix factorization (NMF), in which several constraints are defined to describe the background noise in a generic way. The observed spectrogram is decomposed into four signal contributions: the voiced speech source and three generic types of noise. The speech signal is represented by a source/filter model which captures only voiced speech, and where the filter bases are trained on a database of individual phonemes, resulting in a small dictionary of phoneme envelopes. The three remaining terms represent the background noise as a sum of three different types of noise (smooth noise, impulsive noise and pitched noise), where each type of noise is characterized individually by imposing specific spectro-temporal constraints, based on sparseness and smoothness restrictions. The method was evaluated on the 3rd CHiME Speech Separation and Recognition Challenge development dataset and compared with conventional semi-supervised NMF with sparse activations. Our experiments show that, with a similar number of bases, source/filter modeling of speech in conjunction with the proposed noise constraints produces better separation results than sparse training of speech bases, even though the system is only designed for voiced speech and the results may still not be practical for many applications.

Sinusoidal Modelling with Complex Exponentials for Speech and Audio Signals

In this paper we propose a new approach based on energy-adaptive matching pursuits to improve sinusoidal modelling of speech and audio signals for coding and recognition purposes. To reduce the complexity of the algorithm, an over-complete dictionary composed of complex exponentials is used and an efficient implementation is presented. An analysis-synthesis windows scheme that avoids overlapping is proposed, too. Experimental results show evidence of the advantages of the proposed method for sinusoidal modelling of speech and audio signals compared to some others proposed in the literature.