Simulating Multi-Channel Wind Noise Based on the Corcos Model

Abstract

A novel multi-channel artificial wind noise generator based on a fluid dynamics model, namely the Corcos model, is proposed. In particular, the model is used to approximate the complex coherence function of wind noise signals measured with closely-spaced microphones in the free-field and for time-invariant wind stream direction and speed. Preliminary experiments focus on a spatial analysis of recorded wind noise signals and the validation of the Corcos model for diverse measurement set-ups. Subsequently, the Corcos model is used to synthetically generate wind noise signals exhibiting the desired complex coherence. The multi-channel generator is designed extending an existing single-channel generator to create N mutually uncorrelated signals, while the predefined complex coherence function is obtained exploiting an algorithm developed to generate multi-channel non-stationary noise signals under a complex coherence constraint. Temporal, spectral and spatial characteristics of synthetic signals match with those observed in measured wind noise. The artificial generation overcomes the time-consuming challenge of collecting pure wind noise samples for noise reduction evaluations and provides flexibility in the number of generated signals used in the simulations.

On the Difference-to-sum Power Ratio of Speech and Wind Noise Based on the Corcos Model

“…In the latter case the power ratio of wind noise can be greater than one. The measured data of the wind noise power ratio (solid blue lines) closely follows the theoretical expression (dashed red lines) given by (15). The results shown here were obtained using speech convolved with the direct-path impulse responses and recorded wind noise from the experiment described in [15].…”

“…From (15) it is clear that for larger distances d the exponential decay of the complex coherence of the wind noise γ c (ω k ) rapidly tends toward zero for increasing frequency, so that the wind noise power ratio can be assumed as unitary and frequency independent. Figure 1 shows the comparison between the power ratio of wind noise and the power ratio of clean speech for two different conditions: (a) with a microphone distance of 4 mm, a valuable separation between wind noise and speech is achieved, while for (b) with a microphone distance of 20 mm, the separation is disrupted due to the fact that the speech power ratio increases with frequency for θ s = 90 • .…”

“…The alternation of distorted and clean speech was chosen to include onsets/offsets of wind noise. The speech and wind noise audio items were randomly selected and subsequently mixed from two different databases: the speech signals were selected from the Lib-riSpeech ASR Corpus [18] and wind noise signals were selected from a collection of 100 samples of artificially generated dual-microphone wind noise, from the simulation approach presented in [15]. The dual-microphone clean speech presents no relative-phase difference, simulating a broad-side scenario.…”

“…In this contribution, a definition of the Corcos model is used that holds for every possible direction of the air stream and includes the lateral coherence decay, as shown in [15].…”

Generating coherence-constrained multisensor signals using balanced mixing and spectrally smooth filters

“…In our recent work (Mirabilii and Habets, 2018), we analyzed the noise field recorded with closely spaced microphones that were exposed to the airflow generated by a stationary fan. The low-frequency spatial coherence of the aerodynamic noise agreed well with the empirical model identified by Corcos (Corcos, 1964).…”