INCE:72-Title                                                                                                                                                                
 

ON THE USE OF REAL HEAD RECORDING IN PRODUCT SOUND DESIGN 

Antti Järvinen and Panu Maijala
Laboratory of Acoustics and Audio Signal Processing
Helsinki University of Technology
P.O. Box 3000, FIN-02015 HUT, FINLAND
Tel: +358 9 451 2499 Fax: +358 9 460 224
Email: Antti.Jarvinen@hut.fi
 

The binaural recording technique has been available for a long time. It has been used succesfully for recording material for sound quality analysis and listening tests. The most important reason for the success of the technique is that it offers superior spatial performance compared to other recording techniques, especially if headphone playback is used [1].

Traditionally an artificial head has been used for binaural recording. It is also possible to make binaural recordings using a human subject fitted with microphones in the ears. In this paper this technique will be called the real head recording. Recently it has been shown [2] that the quality of real head recordings made using a randomly selected subject is at least as good as artificial head recordings.

We will present our experiences in binaural recording made using artificial and real heads. Special attention is paid to the practical aspects of the binaural recording technique for product sound design.

The basic idea of the binaural recording technique is to record and reproduce signals to the human hearing accurately. This can be done by replicating the sound pressures at both eardrums. The binaural technique has a long history. For an overview of the technique see [1] [3].
 
So far the use of the artificial head has dominated in binaural recording. One of the main reasons for this is that in many cases recording sessions can be very long and human subjects cannot stay without moving for hours. In product sound design situation is different. Quite often the recording time is short and the recording situation can be reproduced easily.
 
To be perfect binaural recording should be made individually for each person. In most applications this is certainly impossible. In sound quality engineering this could be done for a limited group of subjects and provide more accurate recordings for them. This offers an interesting opportunity to provide better quality recordings for the person who is going to do the sound quality analysis.
 
If a proper equalization is made for recording and playback system it has been shown [4] that authentic localization performance can be preserved in individualized binaural recordings.
 
In the case of nonidividualized real head recording it has been found [5] that if a randomly chosen subject is used for recording, localization performance is significantly reduced. In the same experiments it was shown that if a carefully selected person is used for recording, localization ability is close to the real life situation. Localization ability of artificial head recordings is comparable to randomly choosen human subject [2]. If the localization performance is choosen to be a measure of binaural recording system two conclusions can be made from these results: current artificial heads can be improved and if real head recording technique is used, the head used for recording has significant effect on the quality of recordings.
 
One important but quite often overlooked part of binaural recording is the calibration of the reproduction system. It can be done individually or using a common equalization. It has been found that in the case of carefully designed common equalization, the amount of errors in localization test is only marginally increased compared to individual equalization [5]. If listening tests are considered it should not be a problem to use individual headphone equalization for a listening panel, especially if the same panel is used frequently. How should the real head recordings be made? Basically, a recording can be made at three different positions: at the eardrum, at the entrance of an open ear canal or at the entrance of a blocked ear canal.

It has been shown by extensive measurements that the entrance of blocked ear canal is most suitable for binaural recording [6]. Reasons for this are that this recording position includes complete spatial information and minimum amount of individual information. Also, the standard deviation of the measured head-related transfer functions is smaller in the blocked entrance than in other measuring positions.

We have chosen to make real head recordings at the entrance of the blocked ear canal, because it offers also several practical advantages compared to other recording positions.

3.1 Microphone fitting

One of the main practical reasons to use blocked ear canal for recording is the simple mounting of the microphone. It is relatively easy to insert the microphone in an ear plug to the entrance of ear canal even in an outdoor environment. One can imagine that fitting a probe tube close to the eardrum cannot be done by everyone and everywhere. Also the risk of eardrum damage is high if the probe will move due to shock and vibration. Of course a system which would not have these problems could be designed but the size of it would be so large that it would affect the sound field close to and in the ear.
 
One problem in the blocked ear canal technique is that if a typical ear plug (E-A-R or similar) is used some persons have difficulties in fitting them. We learned also that in driving situations microphones can move or drop from the ear plug if the head is turned suddenly. The use of silicon ear plugs can prevent these problems effectively. Fitting of plugs with microphone is easy because the amount of silicon can be changed according to the size of the ear canal. Microphone dropping is also avoided even in most difficult situations. The only limitation is that the ear plug used for blocking the ear canal also prevents normal hearing.
 
Our experience is that the time used for actual recording is much shorter than the time used to select interesting noise situations to be recorded for later analysis and listening tests.

3.2. Recording system

In our recording system we use miniature electret capsules (Sennheiser KE 4-211-2). The same microphone type has been used for example by Møller. We have developed our own battery powered preamplifier for the system. In the case of the cars or mobile work machines it is very important to record not only sound but also the RPM signal. We have interfaced the preamplifier with the control box of the Cortex MK2 artificial head. This arrangement allows us to store the RPM pulse to the least significant bit of AES/EBU signal. Of course this could be made by using a multichannel recording system.
 
Accurate level calibration can also be made in the field using an adaptor to the standard B&K calibrator.
 
So far we have used a DAT-machine for recording. We found out that in many cases DAT is far from an optimum recording media. It is too sensitive to shock and vibration. Currently we are planning to test hard disk or RAM based systems for recording. For a more complete overview of the system see [7].
 
One of the important benefits of our real head recording system is its price. Microphone capsules and the components for the preamplifier cost around 200$.
 

4. REAL HEAD RECORDING IN MOVABLE WORK MACHINE

A typical application of the binaural recording tehnique is the recording of car interior sound for sound quality analysis. Up to four heads are used simultaneously. Several manufactures make this kind of binaural recording systems.
 
One of the most severe limitations in using an artificial head is that it is also a `dummy' head. It cannot drive nor use the gear shift. In the case of a typical passenger car this is not such a big problem. The cabin can seat easily several persons. Also in many cases it is possible to simulate noise situations in a wind tunnel or other laboratory facility.
 
If movable work machines such as a tractor are considered the situation is completely different. Aerodynamic noise is totally nonexisting and motor, gear and hydraulic noise are the principal noise components. Large parts of these components cannot be found if the machine is not moving.
 
The enviroment where these machines are used is not a road. For example in the forest shock and vibrations can cause noise. There are no laboratory facilities to simulate these conditions. If a laboratory can be found the problem is that the machine cannot fit into it or at least the door is too small.
 
The cabin of these machines is small and typically seats only one person. So there is no room for artificial head and driver. Remote control could be used but it is certainly not very practical. In many cases only few machines are manufactured yearly and the machines are made on order. Machines are expensive and no prototypes are built. It means that recordings should be made quickly and no modifications should be made on the machine under study. Using a small battery powered portable real head recording system all real life situations can be recorded.
 
Of course not everybody is able or willing to drive these machines. We have used standard subjects for easy recording situations. For some situations an operator with long experience with the machine under study has to be used. It would be very helpful if some kind of a field test for choosing the most suitable subject for the recording could be introduced.
 
Compared to a typical car the cabin absorption in mobile work machines is much smaller and the driver is often surrounded by windows. Current artificial heads do not simulate absorption of a clothed person too well. This problem is completely avoided by using the real head technique.
 
Quite often in real life situations the machine is inclined. A human tries to correct this inclination but an artificial head does not.
 

5. CONCLUSIONS
 
The real head recording offers an interesting alternative in binaural recording. We have used the real head recording technique for sound quality analysis in situations where the use of an artificial head was impossible. We discovered that real head technique is a good alternative also in many situations where artificial head can be used.
 
Our recording system is affordable and the sound quality of it is comparable to the artificial head systems.
 
The real head technique will not supress artificial heads but gives an affordable opportunity to use binaural recording if use of an artificial head is difficult or too expensive.
 
It could be concluded that the real head technique offers new possibilities in sound quality analysis and widens the scope of binaural recording.
 
6. ACKNOWLEDGEMENT
 
Authors wish to thank The Finnish Work Environment Fund for financial support.
 
7. REFERENCES
 
1. H. Møller, "Fundamentals of binaural technology.", Applied Acoustics, Vol. 36. No. 3/4. 171-218.
    p. (1992)
2. H. Møller et al, "Evaluation of Artificial Heads in Listening Tests." Audio Engineering Society
    102nd Convention, Munich, 1997, preprint 4404.
3. J. Blauert, Spatial Hearing. The psychophysics of human sound localization. Revised edition,
    MIT Press, Cambridge MA, 1997.
4. H. Møller et al, "Binaural Technique: Do We Need Individual Recordings?" Using a Typical
    Human Subject for Binaural Recording" Journal of the Audio Engneering Society, Vol. 44. No. 6.
    451-469. (1996)
5. H. Møller et al, "Using a Typical Human Subject for Binaural Recording" Audio Engneering
    Society 100 th Convention, Copenhagen, 1996, preprint 4157.
6. D. Hammershøi and H. Møller, "Sound transmission to and within the human ear canal." J. Acoust.
    Soc. Am, Vol. 100. No. 1. 408-427. p. (1996)
7. P. Maijala, "Better binaural recordings using the real human head", Proceedings of Inter-Noise
    1997, Budapest, 1997.