A new method for outlier removal in time-delay-based direction of arrival estimates

Abstract

Recently, the need to monitor restricted areas has increased. Acoustics is one of the available key techniques, but there are some restrictions and constraints to consider. In situations with unknown noise and low SNR the performance of time delay based direction of arrival (DOA) estimators collapses rapidly as SNR decreases. Outliers are introduced into estimation results when signals of interest are masked by noise.There exist several methods for compensation of noise induced errors, such as averaging within subarrays, time delay selection or various minimizations. These compensation methods provide an optimum solution with respect to some criteria, but are ineffective against large errors in multiple time delays.In this paper, we present a method for removing outliers caused by errors in time delays. First, we utilize signal propagation speed to measure an error criterion for DOA estimates. Second, estimates with sufficiently large error criterion are identified as outliers and discarded. We use an adaptive threshold to identify outliers.Effectiveness of our method is verified through experiments with simulations and real data. In both cases we are able to identify and discard outliers and thus improve estimation reliability. Results indicate that the given method can be used to gain efficiency and robustness in DOA estimation applications, such as automatic acoustic surveillance of large areas.