The binaural recording technique builds on the principle of placing microphones on a human’s head - or an artificial head. The microphones' position is either just outside each ear canal or at the bottom of the ear canal close to the eardrum. When you reproduce the recording and listen with headphones, the image is very much similar to “being there”.
This article explains using a binaural headset (instead of an artificial head solution), so the binaural sound recording takes place at the entrance of the ear canal of a real person.
The head affects how we hear
We express the physical influence from the head by the so-called Head-Related Transfer Function or HRTF for short. This transfer function expresses how the head, the ears (and also the torso) affects the transmission of an acoustical signal from a sound source to the eardrums. The size and the shape of the head, the size and the shape of the ears and the distance between the ears all contribute to a filtering of the acoustical signal before it reaches the eardrums.
When the sound comes from in front of a person, the influence is symmetrical, meaning that the sound is the same at both ears. However, as soon as the sound moves to one of the sides, the sound becomes different as it reaches each ear. The level, the frequency response and the time of arrival differ. The side of the head closest to the sound source provides a reflection which causes a pressure build-up in the mid-frequency range. On the opposite side of the head, some shadowing effect occurs in the same frequency range.
HRTF is closely related to the individual shape of the human head. Thus binaural recordings sound the best if recorded using your personal HRTF. However, in general, everybody experiences a much higher envelopment of the surrounding sound, even though the HRTF is not their own. This is why the binaural recording technique is worth considering for many purposes.
Much research has been done regarding binaural sound during the last 30 years. The researchers at Aalborg University, Denmark, carried out comprehensive studies in the 90s [1].
In the next figure, you can see two sets of curves (HRTFs). Each curve represents an average across many subjects. The curves on the left are measured with an open ear canal. The curves to the right show the HRTFs from subjects with blocked ear canal.
Figure 1 HRTF of human subjects, open and closed ear canal [1].
Using the a binaural headset is very close to the HRTF measured with a blocked ear canal (to the right).
Soundbites from the play: Good for Nothing
This play was developed by students of the Danish National School of Performing Arts. Nanna-Karina Schleimann created sound design and musical compositions.
In the play, the audience moves around among the actors. The entire theater space works as the stage and minor “incidents” take place here and there. Nanna chose to use binaural microphones for several of the actors and transmit the signals to the audience by using stereo receivers and headphones. In this way, the audience participates in the most intimate way.
The music was played and mixed live on the stage during performances. The sound bites are recordings taken directly from a performance.
“GFN Alex”: One actor is talking; another actor is listening (wearing the binaural microphones). During this scene they both put their heads into a cardboard box. The audience now experiences being in that box as well.
“Composition”: Composition/sound effects for a scene.
“Pose Paranoia”: Composition/sound effects for a scene.
Up-mix from the binaural sound
The binaural technique is designed for playback with headphones. If the binaurally recorded sound is played back by loudspeakers, some correction is needed. The correction is basically an inversed HRTF. As the HRTF changes with direction, it is complicated to design this inverted function. However, a simple filter approximates the ideal correction. The function of the filter is to recreate the timbral balance to match the timbre of the sound reproduced in headphones.
Figure 2 Proposed filter curve for up-mixing of binaural recordings for loudspeaker playback.
In figure 2, the filter curve shown is created by applying Adobe Audition’s parametric equalizer. Most parametric equalizers can create a curve like this one – or close.
Settings:
480 Hz Low shelf, Gain = +2 dB
4 kHz Bell, Gain = -11 dB, Q = 1
8 kHz Bell, Gain = +8 dB, Q = 2
Master gain = 0 dB
So, when a binaural recording is made, simply add the equalizing to both channels. Then the recording is better adapted for loudspeaker reproduction.
HRTFs are individual so that you might need to use other filter compensations than the one shown above.
The curve below is recorded in a very reverberant space so the sound field is rather diffuse. The sound source is pink noise reproduced by a loudspeaker. An omnidirectional microphone (DPA 4060) is recording the diffuse sound at one point for one minute. Then the omnidirectional microphone is substituted by a person wearing a DPA 4560 and the sound is recorded (in two channels) for one minute. An FFT analysis (FFT size = 8K) is carried out. Left and right spectra from the binaural headset recording are averaged. The measured spectral data from the omni is then subtracted from the averaged spectral data from the binaural recording. The curve shows, more or less, the effect of the head (frequency-wise) when in a diffuse sound field.
Putting the headset on another person will provide a slightly different result.
If you are working with the acquisition of sound effects, you can leave the signal un-equalized until you put it in the mix, to keep the signal as clean as possible for as long as possible. Or you can apply a “less aggressive” equalization like shelving.
This is a “mild” compensation curve, a -6 dB shelf from 2.2 kHz.