Matthieu Puigt

Real-Time Multiple Sound Source Localization and Counting Using a Circular Microphone Array

In this work, a multiple sound source localization and counting method is presented, that imposes relaxed sparsity constraints on the source signals. A uniform circular microphone array is used to overcome the ambiguities of linear arrays, however the underlying concepts (sparse component analysis and matching pursuit-based operation on the histogram of estimates) are applicable to any microphone array topology. Our method is based on detecting time-frequency (TF) zones where one source is dominant over the others. Using appropriately selected TF components in these “single-source” zones, the proposed method jointly estimates the number of active sources and their corresponding directions of arrival (DOAs) by applying a matching pursuit-based approach to the histogram of DOA estimates. The method is shown to have excellent performance for DOA estimation and source counting, and to be highly suitable for real-time applications due to its low complexity. Through simulations (in various signal-to-noise ratio conditions and reverberant environments) and real environment experiments, we indicate that our method outperforms other state-of-the-art DOA and source counting methods in terms of accuracy, while being significantly more efficient in terms of computational complexity.

Temporal and time-frequency correlation-based blind source separation methods. Part I: Determined and underdetermined linear instantaneous mixtures

In this paper, we propose two versions of a correlation-based blind source separation (BSS) method. Whereas its basic version operates in the time domain, its extended form is based on the timefrequency (TF) representations of the observed signals and thus applies to much more general conditions. The latter approach consists in identifying the columns of the (permuted scaled) mixing matrix in TF areas where this method detects that a single source occurs. Both the detection and identification stages of this approach use local correlation parameters of the TF transforms of the observed signals. This BSS method, called TIFCORR (for TImeFrequency CORRelation-based BSS), is shown to yield very accurate separation for linear instantaneous mixtures of real speech signals (output SNR’s are above 60 dB).

Real-time multiple sound source localization using a circular microphone array based on single-source confidence measures

We propose a novel real-time adaptative localization approach for multiple sources using a circular array, in order to suppress the localization ambiguities faced with linear arrays, and assuming a weak sound source sparsity which is derived from blind source separation methods. Our proposed method performs very well both in simulations and in real conditions at 50% real-time.

Source counting in real-time sound source localization using a circular microphone array

Recently, we proposed an approach inspired by Sparse Component Analysis for real-time localization of multiple sound sources using a circular microphone array. The method was based on identifying time-frequency zones where only one source is active, reducing the problem to single-source localization for these zones. A histogram of estimated Directions of Arrival (DOAs) was formed and then processed to obtain improved DOA estimates, assuming that the number of sources was known. In this paper, we extend our previous work by proposing three different methods for counting the number of sources by looking for prominent peaks in the derived histogram based on: (a) performing a peak search, (b) processing an LPC-smoothed version of the histogram, (c) employing a matching pursuit-based approach. The third approach is shown to perform very accurately in simulated reverberant conditions and additive noise, and its computational requirements are very small.

Time-frequency blind signal separation: extended methods, performance evaluation for speech sources

Most reported blind source separation (BSS) methods are based on independent component analysis (ICA), which esp. requires the sources to be stationary (and non-Gaussian). Time-frequency (TF) BSS methods avoid these restrictions and are therefore e.g. attractive for speech signals. We first introduce extensions of three types of TF-BSS methods that we recently proposed, and we analyze the relationships between these methods. We then provide a detailed benchmarking of these methods, based on a large number of tests performed with linear instantaneous mixtures of speech signals. This demonstrates the good performance of these methods (SNR typically above 60 dB) and their low sensitivity to the values of their TF parameters.

Validity of the Independence Assumption for the Separation of Instantaneous and Convolutive Mixtures of Speech and Music Sources

In this paper, we study the validity of the assumption that speech source signals exhibit lower dependency and therefore better separability with Independent Component Analysis algorithms than music sources. In particular, we investigate some dependency measures in the temporal and the time-frequency domains, resp. in the framework of instantaneous and convolutive mixtures. Moreover, we test several ICA methods, based on the above dependency measures, on the same source signals. We experimentally show that speech and music sources tend to have the same mean behaviour for excerpt durations above 20 ms, but music signals provide more spread dependency measures and SIR values. Lastly, we experimentally show that Gaussian nonstationary mutual information is better suited to audio signals than mutual information.

Blind Separation methods based on correlation for sparse possibly-correlated images

In this paper, we propose Blind Source Separation (BSS) methods for possibly-correlated images, based on a low sparsity assumption. To satisfy this sparsity condition, one of the versions of our methods applies a wavelet transform to the observed images before performing separation. Another version directly operates in the original spatial domain, when the sources are sparse enough in this domain. Both methods consist in finding, in the considered sparse representation domain, tiny zones where only one source is active. The column of the mixing matrix corresponding to this source is then estimated in this zone. We also propose extensions of these methods, with automated selection of adequate analysis parameters. Various tests show the good performance of these approaches (SIR improvement often higher than 40 dB).

A Time-Frequency CORRelation-Based Blind Source Separation Method for Time-Delayed Mixtures

We propose a time-frequency (TF) blind source separation (BSS) method suited to attenuated and delayed (AD) mixtures, inspired from a method that we previously developed for linear instantaneous mixtures. This approach only requires each of the uncorrelated sources to occur alone in a tiny TF zone, i.e. it sets very limited constraints on the source sparsity and overlap, unlike various previously reported TF-BSS methods. Our approach is based on time-frequency correlation (hence its name AD-TIFCORR). It consists in identifying the columns of the (filtered permuted) mixing matrix in TF zones where it detects that a single source occurs. We thus identify columns of scale coefficients and time shifts. This method is especially suited to non-stationary sources

Blind Calibration of Mobile Sensors Using Informed Nonnegative Matrix Factorization

In this paper, we assume several heterogeneous, geolocalized, and time-stamped sensors to observe an area over time. We also assume that most of them are uncalibrated and we propose a novel formulation of the blind calibration problem as a Nonnegative Matrix Factorization NMF with missing entries. Our proposed approach is generalizing our previous informed and weighted NMF method, which is shown to be accurate for the considered application and to outperform blind calibration based on matrix completion and nonnegative least squares.

Blind mobile sensor calibration using an informed nonnegative matrix factorization with a relaxed rendezvous model

In this paper, we consider the problem of blindly calibrating a mobile sensor network-i.e., determining the gain and the offset of each sensor-from heterogeneous observations on a defined spatial area over time. For that purpose, we previously proposed a blind sensor calibration method based on Weighted Informed Nonnegative Matrix Factorization with missing entries. It required a minimum number of rendezvous-i.e., data sensed by different sensors at almost the same time and place-which might be difficult to satisfy in practice. In this paper we relax the rendezvous requirement by using a sparse decomposition of the signal of interest with respect to a known dictionary. The calibration can thus be performed if sensors share some common support in the dictionary, and provides a consistent performance even if no sensors are in exact rendezvous.