Version 2 of high-efficiency (HE) advanced audio coding (AAC) introduced in MPEG-4 audio (called HE AAC v2 or eAAC+) has received wide attention recently. One of the key tools in HE AAC v2 is parametric stereo (PS) tool, whose decoded stereo audio sometimes exhibits rapid shifts of the spatial locations of instruments. This situation could be avoided if the encoder is properly designed. Inspired by the concept of perceptual evaluation of audio quality (PEAQ), we propose a model to quantitatively evaluate the quality of a PS encoder in terms of the spatial accuracy of the reproduced sound field. Inside the model, the input signal is analyzed by 32 subband filters. The inter-channel time difference (ITD) and inter-cannel intensity difference (IID) of low-frequency subbands are calculated. The ITD and IID values from the original signal are compared with those from the decoded signal to calculate the overall scores. The experimental results show that the model is able to detect localization discrepancy, and the provided scores are correlated with the preference of audiences. Though the resulting scores are not proportional to that from human listeners, the model nevertheless provides a useful tool for evaluating the perceived spatial quality of the decoded audio.
[1]
William A. Yost,et al.
Spatial hearing: The psychophysics of human sound localization, revised edition
,
1998
.
[2]
J. Blauert.
Spatial Hearing: The Psychophysics of Human Sound Localization
,
1983
.
[3]
Jeroen Breebaart,et al.
Parametric Coding of Stereo Audio
,
2005,
EURASIP J. Adv. Signal Process..
[4]
Bill Gardner,et al.
HRTF Measurements of a KEMAR Dummy-Head Microphone
,
1994
.
[5]
A. Spanias,et al.
Perceptual coding of digital audio
,
2000,
Proceedings of the IEEE.
[6]
Shingchern D. You,et al.
Efficient quantization algorithm for real-time MP-3 encoders
,
2008,
Multimedia Tools and Applications.