Robust signal processing techniques for source localization and multisource spatial sound rendering for immersive environments | Posted on:2003-09-28 | Degree:Ph.D | Type:Thesis | University:University of Southern California | Candidate:Georgiou, Panayiotis G | Full Text:PDF | GTID:2468390011479451 | Subject:Engineering | Abstract/Summary: | | In this thesis we investigate signal processing techniques that can find applications in acoustics. The methods under consideration can be separated in the application domain in: (i) robust source localization using microphone arrays and (ii) spatial sound source rendering; and in the theoretical domain as: (i) robust array signal processing for localization using (a) a low computational fractional lower order statistics method and (b) an optimal Maximum Likelihood localization for multiple dependent sources and noise, and (ii) modeling and filter reduction of the whole 3-D space using a state-space approach for spatial rendering.; Previous work in spatial sound reproduction using HRTF modeling has mainly focused on methods that attempt to model each transfer function individually. These methods are generally computationally-complex and cannot be used for real-time spatial rendering of multiple moving sources. We provide an alternative approach, which uses a multiple-input single-output state-space system to create a combined model of the HRTF's for all directions. This exploits the similarities among different HRTF's to achieve a significant reduction in model size with minimum loss of accuracy.; The robust array signal processing methods are based on the theory of alpha-stable distributions and two approaches are investigated. The first is a low computational alternative to second order statistics cross-correlation based methods. It is based on Fractional Lower-Order Statistics, which mitigate the effects of heavy-tailed noise. An improvement in TDE estimation is demonstrated that is up to a factor four better than what can be achieved with second-order statistics.; The second localization approach considers optimal, Maximum Likelihood localization for multiple sources that are dependent and are present in a dependent noise environment. This work aims to better model effects such as reverberation or echo that appear in room acoustics. We use Sub-Gaussian random processes, which are a special case of symmetric alpha-stable distributions. The Sub-Gaussian with alpha = 1—based on the Lévy (alpha = 0.5)—process is derived and used to formulate the ML function. The separable solution is derived for estimation of the statistics and the Directions-of-Arrival. Subsequently simulations are performed and verify the robustness claim of the Sub-Gaussian based method. Additionally two large linear microphone arrays (20 & 41 microphones) are created and show a significant improvement in localization with the proposed method. | Keywords/Search Tags: | Signal processing, Localization, Spatial sound, Method, Rendering, Robust, Source | | Related items |
| |
|