Hierarchical Bandlimitation of Surround Sound

-- A Psychoacoustical Study

EPSRC Project Reference: EP/C527100/1
Start date: 1 October 2005
End date: 30 Spetember 2008

Principal Investigator: Dr Slawek Zielinski
Co-investigator: Dr Francis Rumsey
Research Student: Dr Yu Jiao


Already in cinemas and on DVDs, and to a growing extent on broadcasts and multimedia systems (e.g. computer games), 5.1-channel surround sound is the norm. It demonstrably improves the entertainment value of the programme, provides a greater sense of involvement for the consumer and brings a new dimension to communication applications. That having been said, not all homes can accommodate a full high-fidelity 5.1 surround sound system and not all broadcast or networked transmission systems can handle the data rates involved.

For that reason many engineers try to quantify the perceptual effects of various simplifications and engineering trade-offs in such systems. For example, a very important (and still unanswered) engineering question is how to limit the overall bandwidth of surround sound with minimal losses in audio quality. This question will be addressed in this project.

Moreover, it is proposed to develop a method for "graceful" bandlimitation of surround sound, that it to say, the bandlimitation causing minimal losses in audio quality. In this method original surround sound signals will be decomposed to a set of independent audio components which are ordered "hierarchically" according to their order of importance. Under the restricted transmission conditions, the overall bandwidth of surround sound is "squeezed" to fit the available bandwidth by extremely limiting the bandwidth of the least important audio components. The proposed method is called "Hierarchical Bandlimitation" which is depicted in the following figure.

In order to develop this method a series of psychoacoustical experiments have to be undertaken. A success in this project could potentially lead to improvements in surround audio quality delivered over broadcast networks or the data networks with restricted bandwidth and could lead to more resourceful use of memory or computational power of audio systems (e.g. in computer games), just to mention some possible applications.

Summary of Results

The main aims of the project was to develop a psychoacoustically optimised Hierarchical Bandwidth Limitation algorithm for multichannel sound. The main findings of this research are summarised as follows. Read the listening test results by clicking on corresponding reports.

  1. It was validated that Gerzon's MSBTF transform could be used as an effective preprocessing method prior to reduction of bandwidth to save the overall bandwidth of surround sound without significant quality degradation. Read report 1.
  2. The Karhunen-Lòeve Transform (KLT) was studied and it was found that eigenchannels of KLT are arranged in a hierarchical order according to their Perceptual Importance. Read report 2.
  3. MSBTF transform and KLT was compared in the context of HBL. It was found that the signal dependent hierarchical encoding technique (KLT) performed better than the signal independent technique (MSBTF) for the purpose of bandwidth limitation. Read report 3.
  4. The perceptual effects of the adaptive KLT-based HBL were studied. Read report 4.
  5. The bandwidth allocation strategy in KLT-based HBL was optimised for 3/2 stereo format for two levels of overall bandwidth: 40 and 60 kHz. Read report 5.
  6. An evaluation has been made of the optimised KLT-based HBL algorithm by comparing it with other bandwidth limitation algorithms. The results showed that the KLT-based HBL with the optimal bandwidth allocation strategy was superior compared to the other traditional bandwidth limitation algorithms used for the limitation of bandwidth of multichannel sound. Read report 6.