[Engineering Report] The localization performance of the source localization algorithms degrades in reverberant conditions. The performance of one such localization algorithm, the localization–extraction (LE) algorithm, was measured systematically as a function of the number of reflecting surfaces in a cubical enclosure. Localization was qualitatively measured using a localization plot and quantized using two objective parameters. A broad-band noise burst and a speech signal were used as stimuli. The degradation of the localization performance was monotonic but not uniform with an increase in the number of reflectors. The performance was found to be proportional to the bandwidth of the stimulus. The performance of the LE algorithm was benchmarked against that of a commonly used signal-subspace technique—multiple signal classification (MUSIC). The LE algorithm was less affected by reflections than the MUSIC algorithm. Degradation of the source localization under high reverberation was found to be more severe at low frequencies, which resulted in the detection of a “phantom” source at 0° for the speech signal.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.