Scientists find solution for better sound in 3D

Scientists find solution for better sound in 3DIndividual head models enable users to adjust frequency response to get full effect of 3D sound. Steve Bush
Researchers from Singapore have been looking to improve 3D sound systems by matching them to the user.
3D sound systems seek to give listeners the impression that sound is being generated all around by adding ‘psycho-acoustic' distortions to audio from conventional stereo loudspeakers. Frequency bands shift sound f₁   f₂   f₃   f₄   f₅   f₆ 225Hz   680Hz   2kHz   6.3kHz   10.9kHz*   22kHz* Band A   B   C   D   E   * Variable
To make a sound appear in front of the listener, bands A, C and E are boosted, while B and D are attenuated, opposite adjustments position a sound to the rear. To increase perceived elevation, the frequency of band breaks f4 and f5 are shifted upwards, while maintaining their frequency difference. Down shifts reduce elevation.  
The clues are generated with knowledge of the way our brains and ears process and modify sound, and a geometric model of the human head.
Each system uses a proprietary model, an average head with average ears, based on research from target markets.
Using these systems, most people get some form of 3D sound experience, but very few - only those that physically match the model - get the full effect.
To achieve this maximum 3D sound experience requires that the psychoacoustic processing uses individual head and ear models for each user. This has so-far proved unacceptably complex and time consuming to implement.
In an article published in the IEE's Electronic Letters (Volume 34), Chong-Jin Tan and Woon-Seng Gan of Nanyang Technological University, have proposed and tested a compromise solution.
It allows the listener to select from a set of stored head models to find the one closest to their own, then adjust the frequency response to compensate for differences between their ears and the average model.   3D sound from two loudspeakers?
The ears and brain detect the direction of sounds using three parameters: sound level difference between the ears, reception time difference between the ears and received signal frequency spectrum compared with known sounds. The brain uses the first two effects to give the angle to the sound source in the horizontal plane, but these clues only start to resolve front-back ambiguity, and neither can be used to extract elevation information. To sort these out, the brain relies on spectral modifications to sounds made by the shape of an individuals ears. California-based SRS Labs license 3D sound systems, as does the UK's CRL with its Sensaura Digital Ear technology.
The first step refines the user's perception of angular displacement in the horizontal plane, the second gives the best fit for adding elevation perception to the sound and solving front-back ambiguity.
The researchers have done much work on the frequency response effects, these being the most difficult to achieve accurately due to the subtle nature of the clues the brain extracts.
The frequency response approach Tan and Gan employ is not new, involving splitting sound into five variable-width frequency bands (see table) and boosting or attenuating each band by 8dB (more at high frequency).
What is new is that the researchers allow the listener to vary the magnitude of the shifts and amplitude changes.
Their results show that the frequency response technique is not perfect, but user-modification significantly improves the number of people that perceive movement.
Of ten subjects, the number experiencing front-back confusion dropped from eight to four when they were able to set their own values and seven got a sense of change in elevation.