Reduced complexity adaptive beamformers

Mittal, Manan

Permalink

https://hdl.handle.net/2142/122078 Copy

Description

Title

Reduced complexity adaptive beamformers

Author(s)

Mittal, Manan

Issue Date

2023-12-08

Director of Research (if dissertation) or Advisor (if thesis)

Singer, Andrew C

Department of Study

Electrical & Computer Eng

Discipline

Electrical & Computer Engr

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• Compressed Beamformer
• Sensor Arrays
• Random Projections
• Beamforming
• Source Separation
• Distributed Arrays

Language

eng

Abstract

In many applications of interest, one may benefit from the use of large sensor arrays to capture the desired signals at multiple spatial locations. However, in complex acoustic environments, signal-dependent approaches like matched filtering have high computational requirements, which can be impractical for real-time applications. In many commercially available systems, the real-time processing requirement restricts the system designers to use signal-independent beamformers, such as a delay and sum beamformer. This work explores the design of hybrid beamforming systems that first use signal-independent beamformers to project the sensor outputs into a space of lower dimension and in turn uses those preprocessed signals to perform signal-dependent beamforming. These systems easily generalize to distributed arrays and can be constrained to meet bandwidth and latency requirements over a network, as well. This work also investigates the use of universal algorithms in order to find a mixture of projections to maximise the signal-to-noise-ratio of the desired signal. The performance of hybrid systems is evaluated in simulation and with real world data. This thesis proposes a system for source separation in reverberant acoustic environments and hypothesizes configurations for beamspace preprocessors. We then extend the system to environments involving a moving listener by designing a hybrid generalized sidelobe canceler and proposing an energy-based beam mixing strategy to combine the outputs of multiple beamformers, which can be identified as an instantaneous Wiener combination. Further, we propose the use of a universal algorithm, in combining the outputs of several hybrid beamformers that can perform nearly as well as the full-complexity beamformer. We then propose using a performance-weighted blend of random preprocessors that is shown to approach the performance of the sensorspace variant over Monte-Carlo trials in a simulated environment. Next, main lobe protection is incorporated into the random projections to derive a universal compressed beamformer that is able to consistently achieve a lower output power than the full-complexity variant in line with Capon's criterion.

Graduation Semester

2023-12

Type of Resource

Text

Handle URL

https://hdl.handle.net/2142/122078

Copyright and License Information

Copyright 2023 Manan Sanjeev Mittal

Owning Collections