Paper Archive

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, a novel approach for the computation of a pitch salience function is presented. The aim of a pitch (considered here as synonym for fundamental frequency) salience function is to estimate the relevance of the most salient musical pitches that are present in a certain audio excerpt. Such a function is used in numerous Music Information Retrieval (MIR) tasks such as pitch, multiple-pitch estimation, melody extraction and audio features computation (such as chroma or Pitch Class Profiles). In order to compute the salience of a pitch candidate f , the classical approach uses the weighted sum of the energy of the short time spectrum at its integer multiples frequencies hf . In the present work, we propose a different approach which does not rely on energy but only on frequency location. For this, we first estimate the peaks of the short time spectrum. From the frequency location of these peaks, we evaluate the likelihood that each peak is an harmonic of a given fundamental frequency. The specificity of our method is to use as likelihood the deviation of the harmonic frequency locations from the pitch locations of the equal tempered scale. This is used to create a theoretical sequence of deviations which is then compared to an observed one. The proposed method is then evaluated for a task of multiple-pitch estimation using the MAPS test-set.

Download

In this paper, we propose the Modulation Scale Spectrum as an extension of the Modulation Spectrum through the Scale domain. The Modulation Spectrum expresses the evolution over time of the amplitude content of various frequency bands by a second Fourier Transform. While its use has been proven for many applications, it is not scale-invariant. Because of this, we propose the use of the Scale Transform instead of the second Fourier Transform. The Scale Transform is a special case of the Mellin Transform. Among its properties is "scale-invariance". This implies that two timestretched version of a same music track will have (almost) the same Scale Spectrum. Our proposed Modulation Scale Spectrum therefore inherits from this property while describing frequency content evolution over time. We then propose a specific implementation of the Modulation Scale Spectrum in order to represent rhythm content. This representation is therefore tempo-independent. We evaluate the ability of this representation to catch rhythm characteristics on a classification task. We demonstrate that for this task our proposed representation largely exceeds results obtained so far while being highly tempo-independent.

Download

Swing is a typical long-short rhythmical pattern that is mostly present in jazz music. In this article, we propose an algorithm to automatically estimate how much a track, a frame of a track, is swinging. We denote this by swing ratio. The algorithm we propose is based on the analysis of the auto-correlation of the onset energy function of the audio signal and a simple set of rules. For the purpose of the evaluation of this algorithm, we propose and share the “GTZAN-rhythm” test-set, which is an extension of a well-known test-set by adding annotations of the whole rhythmical structure (downbeat, beat and eight-note positions). We test our algorithm for two tasks: detecting tracks with or without swing, and estimating the amount of swing. Our algorithm achieves 91% mean recall. Finally we use our annotations to study the relationship between the swing ratio and the tempo (study the common belief that swing ratio decreases linearly with the tempo) and the musicians. How much and how to swing is never written on scores, and is therefore something to be learned by the jazzstudents mostly by listening. Our algorithm could be useful for jazz student who wants to learn what is swing.

Download

In this paper we investigate the problem of singer identification on acapella recordings of isolated notes. Most of studies on singer identification describe the content of signals of singing voice with features related to the timbre (such as MFCC or LPC). These features aim to describe the behavior of frequencies at a given instant of time (local features). In this paper, we propose to describe sung tone with the temporal variations of the fundamental frequency (and its harmonics) of the note. The periodic and continuous variations of the frequency trajectories are analyzed on the whole note and the features obtained reflect expressive and intonative elements of singing such as vibrato, tremolo and portamento. The experiments, conducted on two distinct data-sets (lyric and pop-rock singers), prove that the new set of features capture a part of the singer identity. However, these features are less accurate than timbre-based features. We propose to increase the recognition rate of singer identification by combining information conveyed by local and global description of notes. The proposed method, that shows good results, can be adapted for classification problem involving a large number of classes, or to combine classifications with different levels of performance.

Download