Advanced Applications for Multichannel Teleconferencing Audio Systems

This thesis deals with advanced digital signal processing techniques applied to multichannel teleconferencing audio systems, focusing on their benefits and drawbacks. A teleconferencing system should provide a realistic representation of visual and sound fields, allowing a natural communication among participants anywhere in the world as they were all in the same room. In this context, the work faces several issues mainly concerning on multichannel acoustic echo cancellation. Furthermore, immersive audio reproduction and listening environment equalization are taken into consideration with the aim of improving the performance of conventional teleconferencing systems. All the novel contributions given in the aforementioned scenario are discussed in this thesis. As regards multichannel acoustic echo cancellation, new solutions for channels decorrelation are described [1][2][3][4][5]: they are especially conceived for stereophonic acoustic echo cancellation and based on the “missingfundamental” phenomenon. Moreover, a novel variable step-size approach for improving the convergence speed of adaptive algorithms applied to a stereophonic acoustic echo canceller is discussed [6]. Afterwards, in the field of immersive audio, efficient solutions for real time implementation of Wave Field Synthesis (WFS) [7][8][9] and two possible phase approximations of the driving functions for further lowering the computational complexity are given [10]. Doctoral School on Engineering Sciences Università Politecnica delle Marche Moreover, a WFS-based digital pointing of line arrays is discussed centered on the idea of reproducing a virtual source through WFS formulas [8][11]. With the aim of applying WFS technique to real scenarios, a WOLA-based approach for Wave Domain Adaptive Filtering is discussed [12]. As regards listening environment equalization, the analysis of a novel multipoint magnitude equalization is presented [13][14]. Then, a new solution for phase equalization is introduced in order to improve the perceived audio quality [15]. Eventually, a minor contribution on an ESPRIT-based approach for bass enhancement is presented based on a high resolution spectral estimation technique and a subspace fundamentals tracker [16].