The gloved handclap: an unexpected revolution for simple acoustic measurements?

This article describes the discovery of a new handclap, and the application of it for room acoustic measurements in an ordinary room, a vacant office space without sound absorption. An inevitable disaster? I beg your pardon!

It all started a year ago when I started investigating ‘less accurate’ ways of doing acoustic measurement surveys. Why would anyone deliberately choose less accurate methods? There are several reasons thinkable. No ‘need’ for precision; portability of equipment during (air) transport could be challenging; the location doesn’t have power available; measurements have to be carried out super-fast; etc. etc.

‘Old-school’ measurements typically involve excitation of the room with short-duration pulses. These could be generated by, for example but not limited to, blank pistols, fire crackers, balloons, or spark-trains. I wrote a small paper about impulse response measurements using blank pistols last year [1]. The challenge with blank pistols is that they look (and sound) like a real firearm, which excludes its transportation by public transit. Carrying loudspeakers might be allowed but greatly reduces convenience. Those omnidirectional loudspeakers (including case and power amplifier) easily exceed 20 kg.

We have to go back in time a fit further. Handclaps turn out to be convenient to generate a short pulsive sound. And directly after the ‘clap’ one can listen to the room’s ‘answer’. Recent research by Papadakis and colleagues [2] demonstrated that handclaps can be used for measurement of reverberation time. Compared to the precision method, which involves one of these heavy polyhedron loudspeakers with omnidirectional sensitivity, the deviation in results is within 15% (3 steps of JND) for a signal-to-noise ratio above 35 dB. For survey measurements, which in definition require less accuracy, this could be acceptable. Papadakis and colleagues even teach us for free what handclap configuration has the best potential. I say “for free”, because their open access article has been published under the terms and conditions of the Creative Commons Attribution so everyone can learn from it.

But now I’m going to take it a step further…

DISCLAIMER: Please bear in mind (again) that the motivation of this article is NOT to undermine state of the art engineering methods for room acoustic measurements. Not at all! But, an acoustic consultant must know how much accuracy is appropriate for the occasion of measurements. In my opinion it doesn’t hurt to know the strengths and weaknesses of ‘less-perfect’ measurement methods. Particularly if a ‘less-perfect’ method could someday be the only option.

The discovery of a superior handclap

Last December, I was walking back home from work. The temperature in Beijing wasn’t particularly low, but it was cold enough for me to wear a pair of gloves. They help blocking the ice-cold eastern wind. Like I mentioned earlier in this article, my mind was occupied with ‘less accurate’ measurement. Occasionally I would think of new ways to explore practical issues regarding this.

Some years back, I have done research on echolocation. The outcomes of one of the experiments with blind echolocation experts favoured a signal with a broadband frequency spectrum over narrow-band (tongue click-like) signals. A signal with a broadband spectral content ‘gives back’ more cues in the echo. When it comes to handclaps, it is known that their frequency content is limited to frequencies from 500 Hz and higher. The ‘low-end’ of the handclap sound is constrained by physical aspects of the hands and clapping technique [2]. Among acousticians there seems to be an understanding that handclaps are ‘of limited use’ for acoustic measurements.

During that walk home I passed by a wall, which was about 15 meters away. I clapped my gloved hands to test if I could hear the echo, and I could. I was wearing a pair of leather gloves with fleece lining, see photo 1. The sound of my handclap gave me ideas…

I noticed the clap had more energy in the lower frequencies than a handclap with bare hands. The difference was very convincing. But only the low-frequency sound seemed to differ from a regular handclap. Next, I realized that producing consistent sound (regular) handclaps always seemed difficult to me. So, to say, there’s chance of failure if one doesn’t use a consistent technique (and practices regularly, maybe?). I followed-up with more ‘gloved handclaps.’ And they appeared consistent in sound every time. This had potential. I bonded some more with that clapping sound while I continued my walk towards home.

And then… nothing happened for months. However, when I stumbled on that research paper a few weeks ago [2], the idea came to life again!

Photo 1. Leather gloves with three well-known acoustics handbooks for scale.

Testing in a vacant office space without acoustic absorption

After I found out handclaps are hot again, I decided to do preliminary acoustic testing in a vacant office space, see photo 2. The space had no acoustic absorption, except for a few curtains, a ~15 mm painted foam around the FCU duct, and a pair of metal-stud single-layer gypsum board walls without mineral wool in the cavity. To the ears it sounded reverberant, definitely, but not too much.

Photo 2. Panoramic view of the vacant office space where acoustic testing was carried out.

The test procedure was as follows. An omnidirectional microphone was placed at 1.25 m height above the floor, more than 1 m away from the walls. At 1 m distance, at the same height, handclaps were produced by an untrained individual (myself). 50 handclap attempts (bare hands) with the (recommended) A1+ configuration from [2] where recorded with DIRAC software. Another 50 handclaps with wearing gloves were recorded subsequently at the exact same position. The background noise level was measured too. Lastly, more ‘gloved handclaps’ -impulse responses- were measured according to [3] at 2 fixed receiver positions and 6 random source positions per receiver position. Source-receiver distances were always more than 2 m. I faced the microphone for every measurement. Each measurement was 2.7 seconds in duration.

The following results are analysed:

• Equivalent sound pressure level at 1 m distance for 2 different handclap types. For this a time window of 500 ms was analysed;
• Global A-weighted background noise level in the space;
• Quality of impulse response measurements, described with impulse-to-noise ratio (INR);
• Reverberation time;
• Early-to-late ratio / clarity (C50), compared with theoretical value using Barron’s revised diffuse field theory [4].

Results and discussion

The results from the measurements at 1 m distance are given in Figure 1. A striking difference can be seen in the frequency range 80 – 630 Hz, where the gloved handclap has particularly more energy than the bare-handclap. For the higher frequencies there’s hardly difference. The total A-weighted sound level is similar, but a bit higher for the gloved handclaps. Also, in general, the gloved handclap shows less variation over repeated measurements. The global A-weighted background noise in the space was 40 dB.

Figure 1. Results of equivalent sound pressure level (500 ms time window) at 1 m distance. Error bars represent the standard deviation over 50 measurements. * indicates the 1/3-octave band that has a signal-to-noise ratio (SNR) lower than 10 dB; lower frequency bands all have SNR < 10 dB.

The quality of the impulse response measurements is described using the INR-parameter, which gives the useable decay-range for determination of the reverberation time. INR > 35 dB for determining T20, and INR > 45 dB for T30. As we can see in the results in Figure 2, at 125 Hz not all measurements comply with the requirement for T20. From 250 Hz and higher all measurements have sufficient decay range for T30.

Figure 2. INR of 12 impulse response measurements using gloved handclaps. For T20 an INR > 35 dB is required.

Results for reverberation time are given in Figure 3. It can be observed that a reasonably ‘flat curve’ around 0.9 seconds has been found. The standard deviation shows more variability at lower frequencies. In particular the 125 Hz should be regarded as ‘more uncertain’ due to the inadequate level of the impulse response compared to the background noise. Despite the fact there’s no sound absorption added to the space, a reasonably low reverberation time is measured. I believe there’s a decent amount of diffusion and mixing.

Figure 3. Results of reverberation time, T20 and T30. Error bars represent the standard deviation over 12 measurement positions. Dotted lines are used to mark the shortcoming decay range (INR).

The clarity result is given in Figure 4. The variability on the results is obvious. The standard deviation is often larger than the difference limen (JND = 1 dB). The theoretical result is in line with the spatial-averaged mean (N = 12), despite large variability in results. It is known that source directivity is likely to cause inaccuracies in the early reflected sound, which is of particular significance for the C50 parameter [5]. In [2] the deviations of the (similar) C80-parameter where up to 4 × JND for measurements with sufficient decay range, demonstrating a systematic error regardless of position and resulting SNR.

As mentioned earlier in this article, I was facing the receiver position for every measurement. The handclap could produce more acoustic energy towards the receiver than a uniform omnidirectional source would do. In addition, my torso is likely to reflect some sound towards the receiver position.

Figure 4. Results of clarity (C50). Error bars represent standard deviation over 12 measurement positions. The theoretical result is based on Barron’s revised diffuse sound field theory [4], for which the measured T20 is used.

Conclusions

For reverberation time measurements, the handclap method gives reliable results provided the decay range (INR) meets the requirements from ISO 3382. This article presents in addition to recent research [2] a more useful handclap: the gloved handclap. By wearing leather gloves, the excitation signal will contain significantly more acoustic energy in the lower frequency range than a regular (optimized) handclap has. This directly translates into an improved decay range at lower frequencies in room impulse response measurements. For parameters other than reverberation time, it appears that the handclap method is not appropriate to deliver results that are in line with the precision methods from ISO 3382 [2,3].

Future work

Do we need future work? Yes.

There could be an investigation if the new gloved handclap performs well enough over larger distances, different persons, different gloves, and whether the reverberation time results remain in agreement with precision measurements.

I believe the gloved handclap could be used to quickly assess the room acoustics of an ordinary space. It offers a more balanced frequency spectrum than other impulsive signals. It is friendly to use. Apart from the acoustical benefits, however, it might not look as professional…

References

[1] De Vos, R. Variability in Room Impulse Responses from Survey Measurements using Blank Pistol Shots. Proc. 2019 Audio Technology Exchange Summit, Suzhou.

[2] Nikolaos M. Papadakis, and Georgios E. Stavroulakis. Handclap for Acoustic Measurements: Optimal Application and Limitations. MDPI Acoustics 2020, 2, 224–245.

[3] ISO3382-2(2008) - Acoustics - Measurement of room acoustic parameters - Part 2: Reverberation time in ordinary spaces.

[4] Barron, M. Auditorium Acoustics. E&FN Spon, New York 1993. Appendix B.3, 418-419.

[5] Knüttel, T.; Witew, I.B.; Vorl?nder, M. Influence of “omnidirectional” loudspeaker directivity on measured room impulse responses. J. Acoust. Soc. Am. 2013, 134, 3654–3662.