Informed Array Processing
The concept of informed array processing is introduced in this chapter. The conceptual aim of informed array processing is to incorporate relevant spatial information about the problem to be solved into the design of spatial filters and into the estimatio
- PDF / 2,162,162 Bytes
- 34 Pages / 439.37 x 666.142 pts Page_size
- 89 Downloads / 211 Views
Informed Array Processing
Classical beamformers allow us to control the tradeoff between noise reduction and speech distortion, but are not very robust to estimation errors and source position changes and have a slow response time. In contrast, parametric spatial filtering techniques have a fast response time and are relatively robust, but do not allow us to control this tradeoff, can suffer from audible artefacts when the parametric model is violated, and have relatively poor interference reduction. Informed array processing aims to bridge the gap between these two approaches. The conceptual aim of informed array processing is to incorporate relevant information about the problem to be solved into the design of spatial filters and into the estimation of the second-order statistics that are required to implement these filters. The information that can be used to inform the design of the filter weights and the statistical estimation includes time- and frequency-dependent • signal-to-diffuse ratio (SDR) estimates (obtained using the algorithms in Sect. 5.2, for example); • direction of arrival (DOA) estimates (obtained using the algorithms in Sect. 5.1, for example); • interaural time difference (ITD) or interaural level difference (ILD) estimates; and • position estimates (this typically requires multiple arrays). The informed array processing approach is illustrated in the form of a block diagram in Fig. 9.1, where an informed spatial filter is applied to the spherical harmonic domain (SHD) pressure signals, the eigenbeams, to obtain an enhanced output signal. The eigenbeams are also used to estimate acoustic parameters, which are then used to estimate second-order statistics and (optionally) compute the weights of the informed spatial filter.
© Springer International Publishing Switzerland 2017 D. Jarrett et al., Theory and Applications of Spherical Microphone Array Processing, Springer Topics in Signal Processing 9, DOI 10.1007/978-3-319-42211-4_9
151
152
Eigenbeams
9 Informed Array Processing
Informed Spatial Filter
Filter Output Signal
Second-Order Statistics Estimation Parameter Estimation (direction-of-arrival, signalto-diffuse ratio, etc.)
Fig. 9.1 Block diagram of an informed spatial filtering approach
This approach can be applied to problems such as noise reduction, dereverberation or source extraction. In this chapter, we look at two application scenarios: coherent and incoherent noise reduction (Sect. 9.1) using instantaneous DOA estimates, and joint dereverberation and incoherent noise reduction using instantaneous SDR estimates (Sect. 9.2).
9.1 Noise Reduction Using Narrowband DOA Estimates The implementation of the SHD signal-dependent beamformers presented in Chap. 7 requires the estimation of the second-order statistics of the desired and noise signals, most importantly the power spectral density (PSD) matrix of the noise. Unfortunately, in practice this is not a straightforward problem since the desired and noise signals cannot be observed directly, and their statistics must instead be
Data Loading...
