Paper Archive

Artificial reverberation can be produced either by convolving a signal with a measured room impulse response (RIR) or by synthesizing it with a parametric algorithm such as a Feedback Delay Network (FDN). The former reproduces a captured space faithfully but is costly to run and offers no control over its acoustic properties, while the latter is efficient and editable but hard to match to a specific room. In this paper we bridge the two by fitting a fully differentiable FDN to a measured RIR through gradient descent. The proposed network uses sixteen delay lines at a sampling rate of 48 kHz and trains all of its components jointly, including the delay lengths, the feedback matrix, the early-reflection taps, and a set of attenuation filters that control the frequency-dependent decay.

Download

This work analyzes CLAP audio embeddings through a probing framework, studying the encoding of reverberation (RT60), loudness (LUFS), spectral content (SC), and relative pitch (RP). Results show that all attributes are reliably recoverable from CLAP embeddings, with RT60, LUFS, and RP approximately linearly encoded, while SC requires non-linear probes. The identified patterns generalize across eight additional audio foundation models.

Download

IRCAM has a long experience in analysis, synthesis and transformation of voice. Natural voice transformations are of great interest for many applications and can be combine with text-to-speech system, leading to a powerful creation tool. We present research conducted at IRCAM on voice transformations for the last few years. Transformations can be achieved in a global way by modifying pitch, spectral envelope, durations etc. While it sacrifices the possibility to attain a specific target voice, the approach allows the production of new voices of a high degree of naturalness with different gender and age, modified vocal quality, or another speech style. These transformations can be applied in realtime using ircamTools TR A X.Transformation can also be done in a more specific way in order to transform a voice towards the voice of a target speaker. Finally, we present some recent research on the transformation of expressivity.

Download

In this paper, we present TIV.lib, an open-source library for the content-based tonal description of musical audio signals. Its main novelty relies on the perceptually-inspired Tonal Interval Vector space based on the Discrete Fourier transform, from which multiple instantaneous and global representations, descriptors and metrics are computed—e.g., harmonic change, dissonance, diatonicity, and musical key. The library is cross-platform, implemented in Python and the graphical programming language Pure Data, and can be used in both online and offline scenarios. Of note is its potential for enhanced Music Information Retrieval, where tonal descriptors sit at the core of numerous methods and applications.

Download

Identifying chords and related musical attributes from digital audio has proven a long-standing problem spanning many decades of research. A robust identification may facilitate automatic transcription, semantic indexing, polyphonic source separation and other emerging applications. To this end, we develop a Bayesian inference engine operating on single-frame STFT peaks. Peak likelihoods conditional on pitch component information are evaluated by an MCMC approach accounting for overlapping harmonics as well as undetected/spurious peaks, thus facilitating operation in noisy environments at very low computational cost. Our inference engine evaluates posterior probabilities of musical attributes such as root, chroma (including inversion), octave and tuning, given STFT peak frequency and amplitude observations. The resultant posteriors become highly concentrated around the correct attributes, as demonstrated using 227 ms piano recordings with −10 dB additive white Gaussian noise.

Download

We present the Shift-Invariant Hierarchical Dirichlet Process (SIHDP), a nonparametric Bayesian model for modeling multiple songs in terms of a shared vocabulary of latent sound sources. The SIHDP is an extension of the Hierarchical Dirichlet Process (HDP) that explicitly models the times at which each latent component appears in each song. This extension allows us to model how sound sources evolve over time, which is critical to the human ability to recognize and interpret sounds. To make inference on large datasets possible, we develop an exact distributed Gibbs sampling algorithm to do posterior inference. We evaluate the SIHDP’s ability to model audio using a dataset of real popular music, and measure its ability to accurately find patterns in music using a set of synthesized drum loops. Ultimately, our model produces a rich representation of a set of songs consisting of a set of short sound sources and when they appear in each song.

Download

In this paper we present graphical tools and parameters search algorithms for the timbre space exploration and design of complex sounds generated by physical modeling synthesis. The tools are built around a sparse representation of sounds based on Gammatone functions and provide the designer with both a graphical and an auditory insight. The auditory representation of a number of reference sounds, located as landmarks in a 2D sound design space, provides the designer with an effective aid to direct his search for new sounds. The sonic landmarks can either be synthetic sounds chosen by the user or be automatically derived by using clever parameter search and clustering algorithms. The proposed probabilistic method in this paper makes use of the sparse representations to model the distance between sparsely represented sounds. A subsequent optimization model minimizes those distances to estimate the optimal parameters, which generate the landmark sounds on the given auditory landscape.

Download

This paper proposes a novel method for separating harmonic and percussive sounds in single-channel music recordings. Standard non-negative matrix factorization (NMF) is used to obtain the activations of the most representative patterns active in the mixture. The basic idea is to classify automatically those activations that exhibit rhythmic and non-rhythmic patterns. We assume that percussive sounds are modeled by those activations that exhibit a rhythmic pattern. However, harmonic and vocal sounds are modeled by those activations that exhibit a less rhythmic pattern. The classification of the harmonic or percussive NMF activations is performed using a recursive process based on successive correlations applied to the activations. Specifically, promising results are obtained when a sound is classified as percussive through the identification of a set of peaks in the output of the fourth correlation. The reason is because harmonic sounds tend to be represented by one valley in a half-cycle waveform at the output of the fourth correlation. Evaluation shows that the proposed method provides competitive results compared to other reference state-of-the-art methods. Some audio examples are available to illustrate the separation performance of the proposed method.

Download

Loopers become more and more popular due to their growing features and capabilities, not only in live performances but also as a rehearsal tool. These effect units record a phrase and play it back in a loop. The start and stop positions of the recording are typically the player’s start and stop taps on a foot switch. However, if these cues are not entered precisely in time, an annoying, audible gap may occur between the repetitions of the phrase. We propose an algorithm that analyzes the recorded phrase and aligns start and stop positions in order to remove audible gaps. Efficiency, accuracy and robustness are achieved by including the phase information of the onset detection function’s STFT within the beat estimation process. Moreover, the proposed algorithm satisfies the response time required for the live application of beat alignment. We show that robustness is achieved for phrases of sparse rhythmic content for which there is still sufficient information to derive underlying beats.

Download

Decomposition of sounds into their sinusoidal, transient, and noise components is an active research topic and a widely-used tool in audio processing. Multiple solutions have been proposed in recent years, using time–frequency representations to identify either horizontal and vertical structures or orientations and anisotropy in the spectrogram of the sound. In this paper, we present SiTraNo: an easy-to-use MATLAB application with a graphic user interface for audio decomposition that enables visualization and access to the sinusoidal, transient, and noise classes, individually. This application allows the user to choose between different well-known separation methods to analyze an input sound file, to instantaneously control and remix its spectral components, and to visually check the quality of the separation, before producing the desired output file. The visualization of common artifacts, such as birdies and dropouts, is demonstrated. This application promotes experimenting with the sound decomposition process by observing the effect of variations for each spectral component on the original sound and by comparing different methods against each other, evaluating the separation quality both audibly and visually. SiTraNo and its source code are available on a companion website and repository.

Download