Geometric calibration of very large microphone arrays in mismatched free field.

Large microphone arrays are an efficient means for source localization thanks to a wide aperture and a great number of sensors. When such arrays are deployed in situ, accurate geometric calibration becomes essential to obtain the microphone positions. In free field, the classic procedures rely on measured Times of Arrival (TOA) or Time Differences of Arrival (TDOA) between the microphones and several controlled sources. However, free field model mismatches, such as reflectors, generate outliers which severely deteriorate the positioning accuracy. This paper introduces a unified framework for robust calibration using TOA or TDOA by exploiting an outlier-aware noise model. Thanks to the largeness of the array, the existing outliers are sparse and can be identified by a Lasso regression. From this, three iterative robust solvers are proposed: (i) for TOA by Robust Multi Dimensional Unfolding, a particular variation of Robust Multi Dimensional Scaling, (ii) for TDOA by data predenoising based on sparse and low-rank matrix decomposition, and (iii) for TDOA by jointly identifying the outliers and the geometry. The relevance of outlier-aware approaches is asserted by numerical and experimental tests. Compared with the baseline least-square approaches, the proposed robust solvers significantly improve the positioning accuracy in a free field mismatched by reflectors.

Bone-conducted sound in a dolphin's mandible: Experimental investigation of elastic waves mediating information on sound source position.

Mammals use binaural or monaural (spectral) cues to localize acoustic sources. While the sensitivity of terrestrial mammals to changes in source elevation is relatively poor, the accuracy achieved by the odontocete cetaceans' biosonar is high, independently of where the source is. Binaural/spectral cues are unlikely to account for this remarkable skill. In this paper, bone-conducted sound in a dolphin's mandible is studied, investigating its possible contribution to sound localization. Experiments are conducted in a water tank by deploying, on the horizontal and median planes of the skull, ultrasound sources that emit synthetic clicks between 45 and 55 kHz. Elastic waves propagating through the mandible are measured at the pan bones and used to localize source positions via either binaural cues or a correlation-based full-waveform algorithm. Exploiting the full waveforms and, most importantly, reverberated coda, it is possible to enhance the accuracy of source localization in the vertical plane and achieve similar resolution of horizontal- vs vertical-plane sources. The results noted in this paper need to be substantiated by further experimental work, accounting for soft tissues and making sure that the data are correctly mediated to the internal ear. If confirmed, the results would favor the idea that dolphin's echolocation skills rely on the capability to analyze the coda of biosonar echoes.

A robust and passive method for geometric calibration of large arrays.

This paper presents a complete strategy for the geometry estimation of large microphone arrays of arbitrary shape. Largeness is intended here in both number of microphones (hundreds) and size (few meters). Such arrays can be used for various applications in open or confined spaces like acoustical imaging, source identification, or speech processing. For so large array systems, measuring the geometry by hand is impractical. Therefore a blind passive method is proposed. It is based on the analysis of the background acoustic noise, supposed to be a diffuse field. The proposed strategy is a two-step process. First the pairwise microphone distances are identified by matching their measured coherence function to the one predicted by the diffuse field theory. Second, a robust multidimensional scaling (MDS) algorithm is adapted and implemented. It takes advantage of local characteristics to reduce the set of distances and infer the geometry of the array. This work is an extension of previous studies, and it overcomes unsolved drawbacks. In particular it deals efficiently with the outliers known to ruin standard MDS algorithms. Experimental proofs of this ability are presented by treating the case of two arrays. They show that the proposed improvements manage large spatial arrays.

A hybrid method for calibrating acoustic arrays.

A method to calibrate the elements of large arrays devoted to underwater applications is presented. The goal is to measure the sensitivity and directivity of the elements over their full bandwidth. The main constraint comes from the bounded geometry of the experimental setups that limits the duration of the time windows available for analyzing the received signals. Using a short wideband pulse is detrimental to obtaining high signal-to-noise ratios. A classical method for handling this problem is time-delay spectrometry (TDS), which is based on the transmission of a linear frequency- modulated signal combined with a sliding frequency filter. An alternative, hybrid method based on the transmission of a sequence of time-frequency-limited signals is proposed. This hybrid method is shown to provide the same spectral density as TDS in the frequency scanning, but the filtering process is quite different. The transmitted signals are designed to take advantage of the coherent sums of the received signals to track the time of flight of the direct paths between the source and the elements. In addition, a fitting process based on the calibration geometry of data acquisition enables the boundaries of the interference-free time windows to be precisely delineated. An example of the application is described.