J. Cent. South Univ. Technol. (2011) 18: 2009-2014

DOI: 10.1007/s11771-011-0935-8

Sound transducer calibration of

ambulatory audiometric system utilizing delta learning rule

KIM Kyeong-seop¹, SHIN Seung-won¹, YOON Tae-ho², LEE Sang-min³, LEE Insung⁴, RYU Keun ho⁵

1. School of Biomedical Engineering, Konkuk University, Chungju, 380-701, Korea;

2. Xiu Solution Co. Ltd., Suwon, 442-832, Korea;

3. Department of Electronic Engineering, Inha University, Incheon, 402-751, Korea;

4. Department of Radio Communication Engineering, Chungbuk National University, Cheongju, 361-763, Korea;

5. Database/Bioinformatics Lab, Chungbuk National University, Cheongju, 361-763, Korea

? Central South University Press and Springer-Verlag Berlin Heidelberg 2011

Abstract:

An efficient calibration algorithm for an ambulatory audiometric test system is proposed. This system utilizes a personal digital assistant (PDA) device to generate the correct sound pressure level (SPL) from an audiometric transducer such as an earphone. The calibrated sound intensities for an audio-logical examination can be obtained in terms of the sound pressure levels of pure-tonal sinusoidal signals in eight-banded frequency ranges (250, 500, 1 000, 2 000, 3 000, 4 000, 6 000 and 8 000 Hz), and with mapping of the input sound pressure levels by the weight coefficients that are tuned by the delta learning rule. With this scheme, the sound intensities, which evoke eight-banded sound pressure levels by 5 dB steps from a minimum of 25 dB to a maximum of 80 dB, can be generated without volume displacement. Consequently, these sound intensities can be utilized to accurately determine the hearing threshold of a subject in the ambulatory audiometric testing environment.

Key words:

calibration; audiometric system; pure-tonal sound; sound pressure level; hearing threshold; delta learning rule；

1 Introduction

During hearing loss progression, it is very important to detect hearing deterioration as early as possible, and the early stage screening requires an appropriate audiometric test system. With this aim, a pure-tonal audiometric testing system that involves the generation of only one sine wave frequency at one time can be used to estimate the hearing threshold level of a subject, representing the acuity of the subject for resolving the lowest possible sound intensity [1-3]. Audiometric tests are usually conducted in a special sound-treated room that excludes background or extraneous noise, reducing it to 25 dB at 125 Hz [4]. However, with this manner of testing, the early detection of hearing impairment decreases because persons who want their hearing tested must report to a testing facility. Hence, due to the difficulties in providing hearing healthcare services and the advances of telemedicine and e-health technology, tele-audiology has been developed to especially evaluate the hearing level of a subject in a real-time mode with controlling the remote audiometer via web services [5-6]; and the potentiality of tele-audiometric system was appreciated [7]. Furthermore, GIVENS et al [8] validates the usage of internet-based audiometric testing method by computing the mean and standard error of auditory thresholds. Also, audiology telemedicine applications including hearing aid fitting module were implemented [9] and otoacoustic emissions testing scheme was performed by using remote computing technology [10]. Concerning infant hearing screenings, the feasibility of telemedicine for audiology was examined [11] and the roles of tele-audiology in hearing healthcare service in sub-Saharan Africa [12] and intercontinental hearing assessment [13] were investigated.

NAKAMURA [14] proposed a simple audiometer, termed as ‘Mobile Audiometer,’ by implementing an audiometric test system that used mobile phone ringing tones and a server computer connected to the internet for distributing program items that concerned audiometric service and patient records. This proposed mobile test scheme claimed that it increased the opportunity for early hearing loss detection because testing could be done at convenient time and place of the user. In an ambulatory audiometric system, a sound transducer, such as an earphone not specially designed for audiometric applications, is inserted into ear of a subject. Consequently, as audiometric pure-tone sound is transmitted into the device, the required calibration of the frequency and acoustic amplitude of the sound intensity is emitted by the earphone. HAN and POULSEN [15] suggested a calibration method for setting the equivalent threshold sound pressure level (SPL, p_t) for Sennheiser HDA 200 earphone and Etymotic Research ER-2 insert earphone in the frequency range from 125 Hz to 16 kHz and TROBEN [16] reported free-field correction values for the calibration of DD 45 audiometric earphone for speech audiometry. The various types of sound transducers such as circum-aural earphone, supra-aural earphone, concha earphone, and insert earphone are chosen for the calibration of audiometric equipment using rectangular-pulse test signals [17]. RUIZ et al [18] discussed the uncertainty of audiometer calibration by testing HAD 200 circum-aural earphones in a frequency range from 125 Hz to 8 kHz. Because the calibration process needed to control the sound pressure level precisely, the specific volume intensities of a pure tone at 1 kHz [19] and eight-banded frequency ranges [20] were applied. ZHONG et al [21] realized automatic calibration of sound level meter based on image recognition algorithm and KIM et al [22] achieved a fast and automatic calibration system for a sound level meter in an anechoic room with assuming the system linearity.

In this work, an ambulatory pure tone audiometric system was proposed with a new calibration method that utilizes delta learning rule [23] with encompassing eight-banded pure sinusoidal signals of varying sound pressure levels, progressing from the maximum to the minimum sound levels by steps of 5 dB. This ambulatory hearing loss test system is devised of a personal digital assistant (PDA), so subjects may perform the hearing test at their own convenience. With this testing method, the test subject inserts an earphone into his or her ear canal and confirms the minimum audible intensity level of the sine waves generated by a sound hardware unit within the PDA. It is very important to make sure that the earphone emits the right amount of intensity to accurately estimate the hearing threshold level. However, due to the discrepancy of gain between a sound generator and a receiver system, the actual sound pressure level produced by the earphone can be different from the one that is originally intended. Therefore, a new sound calibration algorithm specific to this audiometric test system is implemented by adopting the delta learning rule to update the weight coefficients. As a result, the system can map the SPL generated from the PDA-based audiometric system into the correct amount of intensity to be emitted from the earphones.

2 Calibration for sound pressure level

2.1 Tuning sets for calibrating sound pressure level

Figure 1 shows the method for obtaining the tuning sets of SPL by measuring the actual intensity of the pure-tonal signals with a sound level meter. The tone signal amplitude generated from the PDA is adjusted to acquire a desired sound pressure level based on the sound level meter reading. The intensity varies by steps of 5 dB from 25 dB to 80 dB encompassing frequency of 250 Hz, 500 Hz, 1 kHz, 2 kHz, 3 kHz, 4 kHz, 6 kHz, and   8 kHz, respectively.

Fig.1 Acquisition system for acquiring tuning sets of SPL:   (a) PDA-based audiometer (HP iPAQ hx4700); (b) Sound- shield box (330 mm × 360 mm × 330 mm) blocking out exterior sound level and reducing it to 21.1 dB; (c) Earphone (Sennheiser HD 250); (d) Sound level meter (CESVA SC-30, measurement indicator range: 0-137 dB)

Denote the desired SPL emitted by the earphone in terms of the sound level meter reading as S, such as

                                (1)

where i and j are the frequency (250, 500, 1 000, 2 000,  3 000, 4 000, 6 000 and 8 000 Hz) and sound indexes (25, 30, 40, 45, 50, 55, 60, 65, 70, 75 and 80 dB), respectively.

Similarly, let X be the desired amplitude of the tone signal generated from the PDA, such as

                             (2)

Let Y be the sound pressure level mapped by W, the weight coefficient matrix is updated by delta learning rule:

                                (3)

In this sound transducer calibration system, W can be updated by

                      (4)

Also, δ_i is an adjusted parameter by the delta learning rule [24] and W is the weighting matrix in the chosen frequency. To reduce the extent of nonlinear increment as the amplitude of the tone signal increases, Eqs.(1) and (3) are transformed in dB scale as follows:

                  (5)

                     (6)

2.2 Delta learning rule stage

Step 1: For a given i-th frequency, measure 20× log(s_j) by a sound level meter and find the parameters a_i and b_i by minimizing the error

Step 2: Compute .

Step 3: Adjust δ_j from 0.1 to 0.0 by steps of 0.001 and determine δ_j such as

Step 4: Apply the fifth degree polynomial fittings to find the coefficients c_i, d_i, e_i, f_i, g_i and h_i by

Step 5: Formulate the weight coefficients such as

Step 6: Repeat for all the operating frequencies and a weight vector W is defined as

Step 7: Repeat Steps 2-5 for the new input vector and update W.

Step 8: Calculate MSE errorp(n)|) for a given frequency.

2.3 Calibration stage for SPL generation

The calibrated SPL from an earphone, Y_i (i=1, 2, …, 8) can be generated by

        (6)

where  is an outer product [24].

3 Experimental results and discussion

Firstly, audiometric experiments were conducted to compare the expected output intensities from a PDA (HP iPAQ hx4700, 624 MHz CPU, 64 MB), X, with those on a sound level meter, Ψ. The experiments were performed in a sound-treated audio booth (1.85 m × 2.4 m × 3.3 m) that kept out extraneous noises by reducing them to  25.2 dB. To reduce the noise level further, we used an additional sound-shield box (0.33 m × 0.36 m × 0.33 m) placed inside the sound-treated room and managed to reduce and block out background noise to a 21.1 dB level. Pure-tone sounds encompassing eight-banded pure sinusoidal signals (250 Hz, 500 Hz, 1 kHz, 2 kHz, 3 kHz, 4 kHz, 6 kHz and 8 kHz) were used with varying sound pressure levels progressing from a maximum sound level to a minimum level by steps of 5 dB. In this manner, we obtained three sets of training data that included pairs of the desired tone signal amplitudes that were emitted from the PDA, X and the actual sound level meter reading Ψ. As explained in steps 1-3 of the training algorithm, this delta learning fitting scheme was compared with the other polynomial fitting equations for a given S_i described by Eq.(5). Figure 2 shows the results at all operating frequencies and it reveals the fact that the error between S_i and the estimated value mapped by the fitting equations can be minimal when the delta learning rule is applied.

A weight vector W is determined after the delta learning rule algorithm is performed. Table 1 shows the weighting coefficients in terms of the fifth degree polynomial. Finally, the calibrated SPLs from Eq.(7) are generated by an audiometric transducer such as an earphone, and measured by a sound level meter as shown in Fig.1. Figure 3 shows the results of the intensity differences between S_i and the reading values of the sound level meter Y_i for i=250 Hz, 500 Hz, 1 kHz, 2 kHz, 3 kHz, 4 kHz, 6 kHz, and 8 kHz, respectively.

Due to the frequency-gain characteristics of a particular PDA device (HP iPAQ hx4700) used in this experiment, Fig.3 reveals that the smaller number of trials is needed to reach around the target sound pressure level at operating frequencies of 500 Hz, 1 kHz, 2 kHz,  4 kHz and 8 kHz, whereas the more trials are necessary at 250 Hz, 3 kHz and 6 kHz. In either case, we can make sure that the PDA-based pure-tone audiometric system can transfer its SPL into the earphones in the right amount of sound intensity.

Table 1 Weight coefficients tuned by delta learning rule

Fig.2 Errors between S_i and a_i x_i+b_i (A),  (B) and δ x_i (C) for frequencies of 250 Hz (a), 500 Hz (b), 1 kHz (c), 2 kHz (d), 3 kHz (e), 4 kHz (f), 6 kHz (g), and 8 kHz (f)

Fig.3 Intensity difference between S_i and reading values on Y_i: (a) 250 Hz; (b) 500 Hz; (c) 1 kHz; (d) 2 kHz; (e) 3 kHz; (f) 4 kHz;  (g) 6 kHz; (h) 8 kHz (1st trial SPL means sound level meter reading on first trial output by mapping Eq.(7); 2nd and 3rd are reading values on second and third trial, respectively)

4 Conclusions

1) An ambulatory audiometric system is implemented using a PDA-based pure-tone test and an earphone sound receiver, which enables users to easily perform hearing loss tests at their own convenience. However, in this system, the reference SPL can vary with the earphone type, and consequently the hearing threshold can be falsely estimated. Therefore, it is very important to make sure that the sound intensity is conducted to ear of a subject via an earphone sound transducer. For this reason, a new calibration algorithm is proposed, in terms of the frequency and amplitude, of a pure-tone sine wave utilizing a delta learning rule for the updating of weight coefficients.

2) In clinical audiometric test systems, the hearing threshold of a subject is evaluated using step by step 5 dB intensity variations at different frequency levels. Therefore, the discrepancy between the reference and the actual earphone sound pressure level must be minimized and the intensity is maintained near 5 dB.

3) The results of the calibrated sound pressure level tuned by the delta learning rule are present. It can be easily seen that the intensity difference is within the limits of 1 dB at all frequencies. Subsequently, this ambulatory audiometric test system and its sound intensity calibration scheme can be used to accurately estimate the hearing threshold.

Acknowledgements

This work was supported by the grant of the Korean Ministry of Education, Science and Technology (The Regional Core Research Program/Chungbuk BIT Research-Oriented University Consortium).

References

[1] NEWSBY H A, POPELKA G R. Audiology [M]. Englewood Cliffs: Prentice Hall, 1992: 126-174

[2] DEBONIS D A, DONOHUE C R. Survey of audiology [M]. Bosten: Pearson Education, 2004: 77-106.

[3] MARGOLIS R H, MORGAN D E. Automated pure-tine audiometry: An analysis of capacity, need, and benefit [J]. American Journal of Audiology, 2008, 17: 109-113.

[4] KHANDPUR R S. Biomedical instrumentation, technology and applications [M]. New York: McGraw-Hill, 2005: 463-485.

[5] GIVENS G D, ELANGOVAN S. Internet application to tele-audiology–“nothin’ but net” [J]. American Journal of Audiology, 2003, 12: 59-65.

[6] YAO J, WAN Y, GIVENS G D. Using web services to realize remote hearing assessment [J]. Journal of Clinical Monitoring and Computing, 2010, 24: 41-50.

[7] SWANEPOEL D W, CLARK J, KOEKMOER D, HALL III J W, KRUMM M, FERRARI D V, McPHERSON B, OLUSANYA B O, MARS M, RUSSO I, BARAJAS J J. Telehealth in audiology: The need and potential to reach underserved communities [J]. International Journal of Audiology, 2010, 49: 195-202.

[8] GIVENS G D, BLANAROVICH A, MURPHY T, SIMMONS S, BLACH D, ELANGOVAN S. Internet-based tele-audiometry system for the assessment of hearing: a pilot study [J]. Telemedicine Journal and e-Health, 2003, 9(4): 375-378.

[9] KRUMM M. Audiology telemedicine [J]. Journal of Telemedicine and Telecare, 2007, 13: 224-229.

[10] KRUMM M, RIBERA J, KLICH R. Providing basic hearing tests using remote computing technology [J]. Journal of Telemedicine and Telecare, 2007, 13: 406-410.

[11] KRUMM M, HUFFMAN T, DICK K, KLICH R. Telemedicine for audiology screening of infants [J]. Journal of Telemedicine and Telecare, 2008, 14: 102-104.

[12] SWANEPOEL D W, OLUSANYA B O, MARS M. Hearing health-care delivery in sub-Saharan Africa—A role for tele-audiology [J]. Journal of Telemedicine and Telecare, 2010, 16: 53-56.

[13] SWANEPOEL D W, KOEKMOER D, CLARK J. Intercontinental hearing assessment—A study in tele-audiology [J]. Journal of Telemedicine and Telecare, 2010, 16: 248-252.

[14] NAKAMURA N. Development of mobile audiometric test system using mobile phones [C]// IEEE EMBS Asian-Pacific Conference. Kyoto, Japan: 2003: 356-357.

[15] HAN L A, POULSEN T. Equivalent threshold sound pressure levels for Sennheiser HDA 200 earphone and etymotic research ER-2 insert earphone in the frequency range 125 Hz to 16 kHz [J]. Scand Audiol, 1997, 22: 105-112.

[16] TORBEN P. Free-field correction values for interacoustics DD 45 supra-aural audiometric earphones [J]. International Journal of Audiology, 2011, 50(6): 361-366.

[17] RICHTER U, FEDTKE T. Reference zero for the calibration of audiometric equipment using ‘clicks’ as test signals [J]. International Journal of Audiology 2005, 44: 478-487.

[18] RUIZ M, FEUEREISEN B, MACHON D, RECUERO M. Factors contributing to the uncertainty in circumaural earphone calibration for audiometric test [J]. Applied Acoustics, 2005, 66(9): 1033-1048.

[19] CHOI J M, LEE H B, PARK C S, OH S H, PARK K S. PC-based tele-audiometry [J]. Telemedicine and e-Health, 2007, 13(5): 501-508.

[20] LEE D H, CHON K M, LEE I W, RO Y S, KIM J D, KONG S K, PARK S S, WANG S G. Development of web-based digital air-conduction pure tone audiometer with automated masking [J]. Korean J Otorhinolaryngol-Head Neck Surg, 2007, 50: 860-868. (In Korean)

[21] ZHANG B, XU H, SUN Q S. HE L B, NIU F, BAI Y, YANG P. An automatic calibration system for frequency weighting functions of sound level meter [C]// Proceedings of the 2010 IEEE International Conference on Mechatronics and Automation. Xi’an: 2010: 401-405.

[22] KIM Y T, LEE Y B, JHO M J. A fast automatic calibration system for a sound level meter in an anechoic room [J]. Applied Acoustics, 2003, 64: 459-470.

[23] FAUSETT L. Fundamentals of neural networks [M]. Upper Saddle River: Prentice Hall, 1994: 86-88.

[24] BISHOP C M. Pattern Recognition and machine learning [M]. Singapore: Springer, 2006: 250-251.

(Edited by YANG Bing)

Received date: 2011-02-15; Accepted date: 2011-07-30

Corresponding author: KIM Kyeong-seop, Professor, PhD; Tel: +82-43-8403765; E-mail: kyeong@kku.ac.kr

Abstract: An efficient calibration algorithm for an ambulatory audiometric test system is proposed. This system utilizes a personal digital assistant (PDA) device to generate the correct sound pressure level (SPL) from an audiometric transducer such as an earphone. The calibrated sound intensities for an audio-logical examination can be obtained in terms of the sound pressure levels of pure-tonal sinusoidal signals in eight-banded frequency ranges (250, 500, 1 000, 2 000, 3 000, 4 000, 6 000 and 8 000 Hz), and with mapping of the input sound pressure levels by the weight coefficients that are tuned by the delta learning rule. With this scheme, the sound intensities, which evoke eight-banded sound pressure levels by 5 dB steps from a minimum of 25 dB to a maximum of 80 dB, can be generated without volume displacement. Consequently, these sound intensities can be utilized to accurately determine the hearing threshold of a subject in the ambulatory audiometric testing environment.