Virtual acoustics

In an ideal binaural sound recording the exact aural cues of any sound event are preserved, and reproduced in the listener’s ears. The simplest binaural recording is obtained using a dummy head with microphone placed in the two ear canals. Playing the binaural recording over headphones reproduces the listener’s experience of the original sound field.

Binaural listening

Binaural listening can be emulated by measuring the frequency responses at the two ears for a sound source at a given orientation to the head. These responses are known as head-related transfer functions (HRTFs). The two HRTFs can be emulated using digital filters, and a pair of binaural signals produced from a single sound source.

Synthesising binaural sound

Virtual acoustic systems are also able to produce binaural sound using loudspeakers. If a listener hears a binaural signal from two loudspeakers, the binaural image is corrupted since the right loudspeaker signal travels across to the left ear, and the left loudspeaker signal travels across to the right ear. This “cross talk” can be removed using a cross-talk canceller, allowing the right loudspeaker signal to reach only the right ear and the left loudspeaker signal to reach only the left ear.

Binaural sound using loudspeakers and a cross-talk canceller

Ideally, the virtual sound scene would not move as the listener’s head moved. Achieving a stationary sound scene requires a head tracker to determine the listener’s head position. The virtual sound scene can be compensated relatively easily with headphone listening by altering the sound source positions in such a way that the perceived source positions are stationary. With loudspeaker listening, however, the cross talk canceller must also be updated with head position, which is a more complex undertaking.

Virtual acoustic systems differ from surround sound systems in that they aim to reproduce sound only inside the listener’s two ear canals, rather than producing an accurate sound field over space. This reduces the complexity of the reproduction system compared to surround systems, and means that low cost personal spatial sound is most practically implemented using virtual acoustics techniques.

Some challenges to high quality virtual acoustics reproduction are:

1. Each listener has a different head shape, and so binaural sounds recorded using a particular head can produce localisation errors when played back for a listener with a different head. These errors are particularly apparent in “cones of confusion”: those sets of source positions where the interaural delay between the two ears are similar.
2. When using loudspeakers, the cross talk canceller implements an inverse filter (equaliser) matrix, and at certain frequencies, the inversion of the matrix is poorly conditioned, resulting in localisation errors. Improved cross talk cancellers for a single listener can be developed using more than two loudspeakers that produce superior virtual acoustic imaging to two-speaker systems. We are developing a virtual acoustic arrays which has this improved robustness.
3. Head-related transfer functions are typically measured for large distances from the listener. They are thus unable to produce near-field effects, which is one of the most convincing aspects of binaural sound. Our research aims to extend virtual acoustics to the production of near-field effects.

We have developed a virtual acoustics system for the efficient production of binaural sound over headphones or loudspeakers. This system has been licensed by Phitek Ltd for application in their noise cancelling headsets for personal mobile sound.