Multisensor, multichannel, and time-space processing

Abstract

This chapter presents time-frequency (t,f) methods suitable for multichannel signal processing using multisensor and time-space processing methods. The topic is covered in seven sections with relevant cross-referencing.

A brief tutorial review of the topic of multichannel/multisensor (t,f) signal processing describes the extension of (t,f) methods to incorporate the spatial diversity information provided by multisensor recordings; this is illustrated on an application to brain EEG abnormality source localization (Section 8.1). The multichannel data can then be processed with time-frequency distributions (TFDs) for channel estimation and equalization. In blind source separation (BSS) and direction of arrival estimation problems, the (t,f) approach to array signal processing leads to improved spatial resolution and source separation performances. Methods include (t,f) multiple signal classification (MUSIC) and TFD-based BSS (Section 8.2). In sensor array processing, for source localization, TFDs provide a good framework for hypothesis testing, and they allow the optimal detector to be implemented naturally and efficiently (Section 8.3). In underwater acoustics and telecommunications, separation of signal mixtures is traditionally based on methods such as independent component analysis or BSS. These can be formulated using TFDs for dealing with the case when the signals are nonstationary (Section 8.4). In the underdetermined case, the (t,f) formulations, methodologies, and algorithms for BSS are implemented using vector clustering and component extraction (Section 8.5). Then, Section 8.6 describes a method where audio source localization and separation can be improved using multisensor (t,f) analysis. Finally, in Section 8.7, a selection of basic algorithm and MATLAB code is provided so as to allow the reader to easily reproduce the results in this chapter.