Realization of adaptive beamformer on open-source hardware

  • Marija Ratkovic Acoustics Laboratory, Department of Telecommunications, School of Electrical Engineering University of Belgrade
  • Miloš Bjelić Acoustics Laboratory, Department of Telecommunications, School of Electrical Engineering University of Belgrade
Keywords: adaptive beamforming, Bela hardware, microphone array, real time processing, signal extraction

Abstract

Realization of an efficient system for extracting useful audio signals in the presence of interferences is an important
engineering problem in the field of acoustics. This paper proposes a method for extracting sound signals from a specific direction of
incidence. A microphone array consisting of eight microphones was used, coupled with space-time signal processing executed in real
time on hardware. The microphone signals are filtered by an adaptive beamformer, whose coefficients are optimized in real time on
open-source hardware using the Least Mean Square algorithm. This paper describes the characteristics and limitations of the used
microphone array. Testing the system through simulation indicates that extracting narrowband and wideband signals using the
presented adaptive beamforming method is theoretically possible. This paper aims to experimentally evaluate the algorithm under
real-world conditions, as well as identify the limitations of extracting wideband signals. Adaptation of the filter coefficients and filtering
the signals from the microphones was implemented on Bela hardware, specialized for processing audio signals. This experiment
highlights the limitations that arise from extracting a wideband signal using this type of hardware in real conditions. By comparing the
useful and filtered signals in both time and frequency domains, the quality of the filtering was analyzed. Our experiments suggest that
the current hardware specifications are a limiting factor for successful wideband signal filtering in real conditions, as they limit the
maximum possible order of the adaptive filter, which proved to be insufficient.

Published
2024-06-26
Section
Original Research Papers