Noise pollution investigation takes advantage of two common methods of diagnosis: measurement using a Sound Level Meter and acoustical imaging. The former enables a detailed analysis of the surrounding noise spectrum whereas the latter is rather used for source localization. Both approaches complete each other, and merging them into a unique system, working in realtime, would offer new possibilities of dynamic diagnosis. This paper describes the design of a complete system for this purpose: imaging in realtime the acoustic field at different octave bands, with a convenient device. The acoustic field is sampled in time and space using an array of MEMS microphones. This recent technology enables a compact and fully digital design of the system. However, performing realtime imaging with resource-intensive algorithm on a large amount of measured data confronts with a technical challenge. This is overcome by executing the whole process on a Graphic Processing Unit, which has recently become an attractive device for parallel computing.

No abstract

The reconstruction of sound sources by using inverse methods is known to be prone to estimation errors due to measurement noise, model mismatch, and poor conditioning of the inverse problem. This paper introduces a solution to map the estimation errors together with the reconstructed sound sources. From a Bayesian perspective, it initializes a Gibbs sampler with the Bayesian focusing method. The proposed Gibbs sampler is shown to converge within a few iterations, which makes it realistic for practical purposes. It also turns out to be very flexible in various scenarios. One peculiarity is the capability to directly operate on the cross-spectral matrix. Another one is to easily accommodate sparse priors. Eventually, it can also account for uncertainties in the microphone positions, which reinforces the regularization of the inverse problem.