Perceptually enhanced spectral distance metric for head-related transfer function quality prediction

Area of Science:

• Acoustics and Psychoacoustics
• Signal Processing
• Virtual Audio Environments

Background:

• Perceptual evaluation of head-related transfer function (HRTF) processing is time-consuming and complex.
• Current numerical methods, like log-spectral distortion (LSD), have limitations in assessing HRTF perception.
• There is a need for more accurate and efficient numerical metrics for HRTF quality assessment.

Purpose of the Study:

• To investigate the correlation between log-spectral distortion (LSD) and the auditory perception of head-related transfer functions (HRTFs).
• To develop a novel, perceptually enhanced spectral distance metric for predicting HRTF quality.
• To compare the performance of the proposed metric against existing methods.

Main Methods:

• Listening tests were conducted using distorted head-related transfer functions (HRTFs) across 32 spatial directions at six levels of log-spectral distortion (LSD).
• Perceived spatial and timbral attributes of the distorted HRTFs were measured.
• A new spectral distance metric was developed involving spectral analysis, threshold discrimination, feature combination, binaural weighting, and perceptual outcome estimation.

Main Results:

• Log-spectral distortion (LSD) demonstrated limitations in adequately assessing the perceptual performance of head-related transfer functions (HRTFs).
• The proposed perceptually enhanced spectral distance metric showed superior performance in prediction error compared to existing methods.
• The new metric exhibited a higher correlation with actual perceptual results than current assessment methods.

Conclusions:

• The developed perceptually enhanced spectral distance metric effectively predicts head-related transfer function (HRTF) quality.
• This novel metric offers a more accurate and reliable alternative to log-spectral distortion (LSD) for HRTF assessment.
• The method has significant potential for evaluating HRTF modeling and individualization research.