Binaural sound source localization using the frequency diversity of the head-related transfer function

Area of Science:

• Acoustics
• Signal Processing
• Human Auditory Perception

Background:

• Head-related transfer function (HRTF) spectral cues are crucial for human sound source localization.
• Current localization techniques struggle to fully leverage HRTF diversity, leading to errors like front-to-back confusions and poor vertical plane localization.
• Limited measured signals in existing methods exacerbate localization challenges.

Purpose of the Study:

• To evaluate a novel source location estimator designed to exploit the frequency domain diversity of the HRTF.
• To assess the estimator's performance in localizing single and multiple sound sources in both horizontal and vertical planes.
• To address limitations of current methods in accurately determining sound source positions.

Main Methods:

• Developed a method for extracting directional information from subbands of broadband signals.
• Introduced a composite estimator utilizing signal subspace decomposition.
• Experimentally evaluated localization performance across various source scenarios and planes.

Main Results:

• The proposed estimator successfully localized sound sources, demonstrating improved accuracy.
• Effectively resolved ambiguities in the vertical plane, a common challenge for existing techniques.
• Investigated the influence of source location, source knowledge, and reverberation on performance.

Conclusions:

• The novel estimator effectively utilizes HRTF spectral cues for enhanced sound source localization.
• The method shows significant potential for improving accuracy and resolving vertical plane ambiguities.
• Further analysis confirmed the impact of environmental factors and source characteristics on localization performance.