The influence of helmets on sound localisation in motorcyclists

Seán Byrne 1

John Kennedy 2

Trinity College Dublin, the University of Dublin College Green Dublin 2 D02 PN40 Ireland

ABSTRACT Motorcyclists are among the most vulnerable road users globally, they enjoy less protection than car drivers and experience higher speed collisions than cyclists, with less mobility and spatial awareness to avoid crashes. Sound localisation (SL) while wearing military and ski helmets has been a significant area of research to improve spatial awareness, but no research has been completed thus far on the impact of motorcycle helmets on SL. This research aims to examine experimentally the impact of a motorcycle helmet on the Head Related Impulse Response, and therefore determine the extent of the impact on SL ability of a user. A database of HRIRs was established as a function of position in the horizontal plane with and without a motorcycle helmet. The variation in the signals was evaluated and it was found that the helmet results in a significant alteration to the HRIR at higher frequencies, important for resolving front-back confusions. It is also demonstrated that the overall sound attenuation would result in reduced SL ability of the rider. A novel system to compensate for changes in the HRIR was developed and experimentally tested to mitigate the impact on SL ability for the wearer.

1. INTRODUCTION

Motorcyclists are considered by the World Health Organisation (WHO) to be in the vulnerable road users category. Along with the others in this vulnerable group (cyclists and pedestrians) they make up more than 50% of road deaths worldwide [1]. In Ireland, motorcyclists make up less than 1.5% of taxed vehicles on the road, but are heavily over represented in deaths, with seventeen out of one hundred and forty eight road deaths, or 11.5%, in 2020 being motorcyclists [2]. Considering recent statistics as of November 15th 2021, there were one hundred and eighteen road deaths in Ireland and of these 20 were motorcyclists [3]. Considering all of this, any advances which might make motorcyclists safer on the road should be pursued. This project aims to improve the spatial awareness of motorcycle riders by improving their Sound Localisation (SL) ability whilst wearing a helmet.

1 byrnes69@tcd.ie

2 jkenned5@tcd.ie

a slaty. inter.noise 21-24 AUGUST SCOTTISH EVENT CAMPUS O ¥, ? GLASGOW

The human brain uses both temporal and spectral cues for Sound Localisation (SL). A sound approaching the head from any angle other than straight in front or behind will arrive at the left and right ear at di ff erent times. This is a binaural acoustic cue and is known as the Interaural Time Di ff erence (ITD). The main factor impacting the ITD, other than angle of incidence, is the size of the listener’s head. The head being between the two ears also produces a di ff erence in the SPL experienced at each ear, with the ear further from a sound source usually experiencing lower amplitude noise. This is known as the Interaural Level Di ff erence (ILD). The ILD can be extended to consider the level di ff erence in particular frequency bands. Lower frequencies will yield lower ILD than higher frequencies because they are less impacted by the presence of the head, with the reverse being true for higher frequencies. These binaural cues alone are limited in their e ff ectiveness by the phenomenon known as the "Cone of Confusion", whereby for every position in the rear quadrant there is a corresponding point in the front quadrant which will yield the same ILD and ITD [4] [5]. To resolve this, the spectral content of a signal is also examined by the human audio processing systems. The pinna has a significant impact on this component of sound localisation, amplifying certain frequencies above 3kHz and distorting the signal in di ff erent ways depending on the source position in the horizontal and vertical planes [6]. The upper bounds of useful frequencies for SL is not clear from the literature, at higher frequencies small geometric changes to the pinna can lead to large changes in the HRIR. Covering of the pinna, and low pass filtering of noise at 4kHz both resulted in appreciable reduction in SL ability of subjects [7] [8]. It is clear that the pinna are covered when using a motorcycle helmet, and Kennedy et al. noted an increasing insertion loss (IL) with frequency for a helmet in [9], which could have similar e ff ects to low pass filtering of the signal. This paper will expand on these findings further. The objectives of this research were as follows:

1. To validate a test setup for measurement of the influence of the helmet using a low cost test rig [10] and make comparisons to Head Related Impulse Response (HRIR) and Head Related Transfer Function (HRTF) data from the CIPIC database. [11].

2. To measure the insertion loss (IL) of the helmet and make comparisons to literature [9]. Literature results will be expanded by examining the variation in IL with change in helmet angle relative to the sound source. The impact this is likely to have on SL ability is considered.

3. To quantify the impact of a motorcycle helmet on the HRTF of a binaural head. From this, predictions for the impact of the helmet on SL ability of the user are made. This testing was completed with the helmet visor both open and closed, in recognition of the fact that motorcyclists will often open the visor for improved ventilation and vision in low speed urban environments.

4. To create and test a system which could reproduce the natural HRTF at ear for a subject wearing a motorcycle helmet.

2. METHODOLOGY

Testing was completed using a 3D printed binaural head previously validated for measurements in the horizontal plane [10]. The head is placed on a tripod, elevated to 1.2 m, and connected to a turntable. A loudspeaker is placed 1 m from the head and raised to the same height. Measurement of the HRIR at each angle involves taking a 30 second recording with microphones mounted in the ear canals of the head, followed by rotating the head by 5 ◦ . The transfer function between the at ear signal and the source signal is calculated as in equation 1, where ˆ G xy ( f ) and ˆ G xx ( f ) are the cross-spectral density function and the auto-spectral density functions respectively. A single block size of 15 seconds was sent to the FFT function in MATLAB. The impulse response is recovered from the inverse Fourier transform of the frequency response function.

(a) Front view of helmet with surface pinna

(b) Side view of helmet with surface pinna

Figure 1: Images of helmet with surface pinna

H ( f ) = ˆ G xy ( f ) ˆ G xx ( f ) (1)

The calculated impulse response also includes the e ff ects of the room. The impulse response is truncated to only the HRIR according to the procedure outlined in [10]. The microphone data were acquired through a National Instruments DAQ and the signals were recorded in MATLAB. In this case the head and ears were instrumented using GRAS 40PL microphones as they have a frequency response of ± 1 dB within 50 - 5,000 Hz, ± 2 dB within 5 - 20 kHz and upper limit of the dynamic range of 150dB re 20 µ Pa . A Genelec 8020B loudspeaker with frequency response of ± 2.5 dB within 66 - 20,000 Hz was placed 1 meter from the inter-aural axis of the dummy head, in a large, non-anechoic room. A helmet was then mounted on the head and the procedure was repeated. Results were compared to those without the helmet to test its IL and its impact on HRIR and HRTF. This was repeated with the visor open. A system to reproduce the HRTF for the user of a motorcycle helmet involved mounting a pair of surface microphones on the helmet. These surface sensors were GRAS 40Ps microphones with a frequency response of + 1 , − 2 dB within 10 Hz - 12,000 Hz and + 1 , − 6 dB within 10 Hz - 20,000 Hz. These were mounted using adhesive pads, and care was taken to mount them symmetrically by measuring the distances from the microphones to key features on the helmet. A pinna was 3D printed from PLA and placed over the surface microphones to examine the improvement this yielded in HRTF of the system, as shown in figure 1.

3. RESULTS AND DISCUSSION

The IL was measured for 360 ◦ in increments of 5 ◦ by subtracting the third octave band sound pressure of each ear of the bare head from that with the helmet. The results at 0 ◦ and 90 ◦ are presented for the one-third octave bands in Figure 2. Despite being a di ff erent commercial helmet and di ff ering test procedures the results at 0 ◦ closely match those recorded by Kennedy et al. in [9]. IL values can be seen to reach 40dB at frequencies above 4kHz. As discussed in [7], signals that are low- pass-filtered at 4kHz resulted in a significant reduction in the subjects ability to localise the source of that signal. Therefore, these results imply that a user of this motorcycle helmet would experience

100 200 400 800 1600 Frequency (Hz) 3150 6300 112500

(a) Left ear at 0 ◦

(b) Right ear at 0 ◦

(c) Left ear at 90 ◦ (closest to source)

(d) Right ear at 90 ◦ (furthest from source)

Figure 2: Insertion loss of helmet at left and right ear

100 200 400 800 1600 Frequency (Hz) 3150 6300 112500

a reduction in their SL ability. A gain at low frequencies can also be observed, matching results in [9]. It is not clear what causes this amplification, although it is likely the result of resonance in the structure of the helmet. The IL of the helmet changes as a function of angle, such that the ear facing away from the source experiences more amplification of the signal at low frequencies and lower losses at high frequencies. The increase in magnification of the signal could be due to the lower amplitude of the signal recorded at the ear further away from the sound source. The smaller values of IL measured when the ear is facing away from the speaker are expected to be due to the noise floor of the instrumentation being reached. This implies that a stronger source signal should be used to verify the true IL of a motorcycle helmet. The ITD was taken from the HRIR as the di ff erence in arrival time between the peak values of the left and right HRIR. The results of this are presented in figure 3 as a function of angle. It can be seen that the ITD with the helmet matches that without the helmet somewhat, but with an obvious di ff erence in the shapes at ± 120 ◦ , at which point there is a significant discontinuity in the ITD with the helmet. This is expected to be a result of the large cavity in the front of the helmet caused by the visor. When the helmet reaches a critical angle where the distance travelled by the sound to each ear can include a transmission path through the visor and air cavity, a substantial decrease in ITD is observed. This was examined further by repeating the test with the visor open. It was thought that if the visor was causing the sudden change in ITD, it being open would make this clearer. This appears to be the case from the results presented in figure 3. The HRTF was calculated in MATLAB and is presented in figure 4 for the bare head and helmet cases. Since the results at the left and right ear are symmetrical, the ILD for each frequency band can be seen from the graph of HRTF by comparing the result at each angle to its opposite angle. It can be seen in figure 4 that very little of the HRTF survives transmission through the helmet. This is

SPL (dB) 8 g 100 200 400 800 1600 Frequency (Hz) 3150 6300 112500

SPL (dB) 8 g 40 100 200 400 800 1600 Frequency (Hz) 3150 6300 112500

Figure 3: ITD of helmet with visor open and closed compared to bare head case

a consequence the variable insertion loss with frequency and angle, which distorts the signal. It can be seen that the standard deviation in results increases substantially with frequency, with a maximum value of 21dB at 7938 Hz. At 16kHz, this increases further to 28dB. This relationship between decreasing repeatability of results and frequency is to be expected in light of the results for insertion loss presented previously. Similarly, if the noise floor of the instrumentation is reached due to the very high IL of the helmet then the results are no longer an accurate function of angle. The fact that the on head pinna is deformed and covered probably also contributes to the deviation in the results from the bare head case, as discussed in Section 1. It is interesting to note that at certain angles the HRTF at 1102 Hz gets closer to the bare head case than at 441 Hz. This closely matches the frequency in Figure 2 where the IL goes through zero. The point where gain equals zero changes as a function of angle as was seen previously, so if this is the cause of the return of the HRTF shape to the bare head case, it is likely that di ff erent ranges of angle would match the bare head case better for frequencies around 1000 Hz. Even at this point where the gain is zero, however, significant modification of the HRTF is still evident. A system was designed to recover the HRTF by mounting a pair of surface microphones on the outside of the helmet. This system would be capable of reproducing the amplitude of the noise outside the helmet for the user but would not be able to reproduce an ITD equivalent to the bare head case. This is because the helmet is substantially larger than the human head, so the distance between the ears is not comparable. It might be possible to position the microphones on the surface such that the distance between them is comparable to a pair of human ears. However, no position exists where this would be true for both angles in front of and to the rear of the head, an optimal position has not been explored here. It can be seen that with the addition of surface microphones, the amplitude of the HRTF is much closer to results from the bare head case. It was noticed that there was increasing deviation as higher frequencies are tested, especially in the 4kHz region. As discussed previously, this is the frequency range which is particularly a ff ected by the pinna due to the wavelengths at these frequencies being comparable in their dimensions to the features of the pinna. In light of this, a pair of pinna were added to the system to test if this improved

00 240 am

90" 45" 180" 315° 210°

2 508 a

(a) 220 Hz

(b) 441 Hz

(c) 1102 Hz

(d) 1984 Hz

(e) 3969 Hz

(f) 7938 Hz

Figure 4: Left ear HRTF of bare head, helmet and HRTF recreation systems at selected frequencies

o 1008 45" 315° 90" 18° 205° 180" 210°

the results. The results are shown in magenta in figure 4. It can be seen that this addition significantly improved the results for the HRTF at 4000 Hz and has not negatively a ff ected the performance of the system at lower frequencies. At frequencies up to 1100 Hz, the agreement between results is better with the addition of pinna, and the agreement is only slightly worse at 2000 Hz. At very high frequencies of 8000 Hz and above there exists significant deviation between all measurements. Higher frequencies are more a ff ected by small variations in position and geometry. In this experiment, the positioning of the pinna on the helmet is arbitrary so it remains to be seen if the recovery of the natural HTRF can be improved at higher frequencies. There is a geometric step from the surface of the helmet to the pinna, the pinna is at an angle due to design decisions made, the material used was PLA which is not a flexible material similar to a natural pinna. These reasons will all contribute to the cause for the lower performance of the system at higher frequencies. It is unclear from the literature what higher frequencies can be considered unimportant for SL, so it is not possible to say if these deviations at 8000 Hz and above would have a meaningful impact on the SL ability of a user. A clear success of the system is the recovery of the HRTF up to 4000 Hz with reasonable accuracy. This would likely be of great benefit to the SL of the wearer and the spectral content of the noise sources surrounding a motorcyclist on the road is predominately at frequencies below 4000 Hz. It is planned that this system will be tested on a group of subjects to test its validity in a realistic setting.

4. CONCLUSION AND OUTLOOK

The binaural testing rig used in [10] was reproduced and used to evaluate the impact of a motorcycle helmet on sound localisation. After validation the impact of a motorcycle helmet on the HRIR and the HRTF was tested and compared to the same result sets for a model of a bare head. The insertion loss of the helmet was measured for a full rotation in 5 ◦ increments. The results matched those recorded in literature at 0 ◦ [9]. It was found that the values of loss and gain, as well as the frequency ranges, change as a function of angle. It follows that this would have a negative impact on the SL ability of a user. It was found that both binaural and monaural cues used for SL, including the ITD, ILD at select frequencies and the HRTF are significantly impacted by the introduction of a motorcycle helmet. Significant deviations between the bare head and the motorcycle helmet ITD exist, with discontinuities at specific angles that are thought to be a result of the helmet visor cavity. The shape of the HRTF shows poor agreement between the bare head and the helmet cases at higher frequencies. In addition to this the ILD becomes very small and increasingly independent of the direction of the sound source as frequency increases. A system has been designed which recovers the HRTF of the user up to 5kHz, at which point larger deviations between the results of up to 8dB are noted. Better results at higher frequencies using this system are expected to be possible through optimisation of the geometric setup. A better design of the pinna mounting on the surface could be completed, and the pinna could be reprinted in materials with acoustic properties closer to a natural pinna. There is a limit to how closely the results can match however, since the features of the helmet are not the same as those of the head, and will therefore impact the sound field in di ff erent ways no matter how the pinna are designed and mounted. It is noted that most noises which a motorbike rider is interested in localising are likely to be below 5kHz, with a petrol engine producing noise mainly in the 1600Hz to 4000Hz regions [12] and very few noises of interest on the road being in the 7kHz or greater regions. In light of this, this system improves localisation ability of a subject in those frequency regions which are most important. The next stage of the work requires the surface mounted microphone signal to be accurately presented to the ear canal within the helmet. Validation of this presentation could be assessed through the HRIR and HRTF of the entire system, including the ear phones used for presentation of the external audio at ear. Additionally listening test could be performed to evaluate the required level of

accuracy on the reproduced HRIR.

REFERENCES

[1] WHO - Road tra ffi c injuries. https: // www.who.int / news-room / fact-sheets / detail / road-tra ffi c- injuries. Cited 2021-10-17. [2] RSA.ie - Provisional Reviews. https: // rsa.ie / RSA / Road-Safety / RSA-Statistics / Collision- Statistics / Provisional-Reviews / . Cited 2021-10-17. [3] An Garda Siochana – roads fatalities to date for 2021. https: // www.garda.ie / en / roads- policing / statistics / roads-policing-fatalities-to-date-for-2021 / . Cited 2021-11-15. [4] B. G. Shinn-Cunningham, S. Santarelli, and N. Kopco. Tori of confusion: Binaural localization cues for sources within reach of a listener. The Journal of the Acoustical Society of America , 107(3):1627–1636, March 2000. [5] S. Carlile, S. Delaney, and A. Corderoy. The localisation of spectrally restricted sounds by human listeners. Hearing Research , 128(1-2):175–189, February 1999. [6] E. A. G. Shaw and R. Teranishi. Sound Pressure Generated in an External [U+2010] Ear Replica and Real Human Ears by a Nearby Point Source. The Journal of the Acoustical Society of America , 44(1):240–249, July 1968. [7] A. D. Musicant and R. A. Butler. The influence of pinnae [U+2010] based spectral cues on sound localization. The Journal of the Acoustical Society of America , 75(4):1195–1200, April 1984. [8] Suzanne K. Ro ffl er and Robert A. Butler. Factors That Influence the Localization of Sound in the Vertical Plane. The Journal of the Acoustical Society of America , 43(6):1255–1259, June 1968. [9] J. Kennedy, N. Holt, M. Carley, and I. Walker. The Influence of the Acoustic Properties of Motorcycle Helmets on Temporary Hearing Loss in Motorcyclists. Acta Acustica united with Acustica , 100(6):1129–1138, November 2014. [10] D. O’Connor and J. Kennedy. An evaluation of 3D printing for the manufacture of a binaural recording device. Applied Acoustics , 171, January 2021. [11] HRTF Data – The CIPIC Interface Laboratory Home Page. Cited 2021-10-21. [12] Y Liu, Y B Jia, X J Zhang, Z C Liu, Y C Ren, and B Yang. Noise Test and Analysis of Automobile Engine. Applied Mechanics and Materials , 307:196–199, 2013. Conference Name: Mechatronics and Computational Mechanics Publisher: Trans Tech Publications Ltd.