J. Biomedical Science and Engineering, 2010, 3, 861-867 doi:10.4236/jbise.2010.39116 Published Online September 2010 (http://www.SciRP.org/journal/jbise/ JBiSE ). Published Online September 2010 in SciRes. http:// www.scirp.org/journal/jbise Using active echo cancellation to minimize stimulus reverberations during hearing studies conducted with the auditory brain response (ABR) technique Mark R. Patterson1, Andrij Z. Horodysky1, Bruce W. Deffenbaugh2, Richard W. Brill3 1Virginia Institute of Marine Science, College of William & Mary, Gloucester Point, USA; 2Biomechatronics Group, Media Laboratory, Massachusetts Institute of Technology, Cambridge, USA; 3Northeast Fisheries Science Center, National Marine Fisheries Service, NOAA, Woods Hole, USA. Email: rbrill@vims.edu Received 26 March 2010; revised 11 March 2010; accepted 15 March 2010. ABSTRACT Because of the physical properties of water as sound conducting medium and the proximity of tank walls, creating an anechoic environment underwater is both technically difficult and expensive to implement. Con- ducting hearing studies of aquatic animals can there- fore be challenging due to stimulus reverberations. To address this issue, we developed MATLAB scripts capable of pre-compensating acoustic stimuli result- ing in location-specific echo cancellation. Our proce- dures are specifically designed for hearing studies conducted with the auditory brain response (ABR) technique. Broadband white noise is used to charact- erize the system response and the digitized acoustic signal subsequently used to generate an acoustic inv- erse file capable of cancelling reverberations. Echo cancellation is nearly perfect, although location-spe- cific. The effectiveness of echo cancellation dimin- ishes with distance from test subject and hydrophone (or microphone) used to create the pre-compensated signal. This distance must be minimized and should preferably be less than 5 cm. The spectral composi- tion of the sound signal is not greatly affected, how- ever. We have successfully used the procedure during hearing studies of sev eral fish species, includ ing yellow- fin tuna (Thunnus albacares). ABR experiments on the latter were done at sea aboard an oceanographic research vessel, a highly echoic environment. Keywords: Acoustic; Aquatic; FFT; Fish; Noise; Tank 1. INTRODUCTION In studies of hearing in terrestrial organisms, anechoic chambers are often used so that the response of the sub- ject to the source stimulus, and not echoes, can be accu- rately assessed [1]. Creating an anechoic environment underwater is more problematic, however, because of the physical properties of water as sound conducting med- ium [2,3] and the proximity of tank walls. Similar to the situation in air, it is possible to create anechoic chambers using acoustic wedges [4], compliant materials that rap- idly dampen wave propagation [5], piezoelectric ce- ramic with embedded shunted resistors [6], or active techniques involving coatings [7]. New smart materials are also available to dampen echoes and can yield a 17-24 dB reduction [8], but they are hard to obtain. Moreover, all the above implementations are expensive, especially for hearing studies of aquatic organisms. One cost estimate for an anechoic flow-through un- derwater chamber was US$ 330K [9]. As a result, re- searchers interested in either the hearing abilities or the effects of sound on the behavior and health of aquatic organisms [10] are often forced to choose less expen- sive measure s such as placemen t of the test aquariu m in an anechoic room [11]. Even this solution, however, does not eliminate echoes within the tank itself. The generation of microbubble curtains is effective for ul- trasonic frequencies [12], but many aquatic organisms hear only much low er frequen cies [13]. As a low-cost alternative, we developed a dig ital signal proc- essing technique, implemented in MATLAB (Math Works, Natick, Massachusetts, USA), capable of building stimulus waveforms that cancel reverberant echoes. Thus, only a win- dowed tone burst is perceived by the test subject. Originally developed for hearing experiments in fishes, our technique is also applicable in air. The system is readily implemented with an omnidirectional hydrophone (or microphone) connected to an analog-to-digital converter with anti-aliasing low-pass filter, a sound source connected to a digital-to-analog converter with impedance matched amplifier, and software for recording
M. R. Patterson et al. / J. Biomedical Science and Engineering 3 (2010) 861-867 Copyright © 2010 SciRes. JBiSE 862 waveforms from the hydrophone and playing signals back through a sound emitter. We have successfully im- plemented our active echo-cancellation procedure with a signal-processing system (model: System II) from Tuck- er Davis Technologies (Alachus, Florida, USA). Conducting hearing studies within a specialized anec- hoic environment also generally presupposes working with species that can be transported to the laboratory. The ability to access hearing abilities of a broader range of aquatic animals than fit this criteria is becoming in cr- easingly important because of the need to understand and predict the d isruptive ef fects of anthropog enic noise, especially in the marine environment [14,15]. For exam- ple, the proposed expansion of wind-generated electrical power (i.e., “wind farms”) to continental shelf areas appears likely to increase ambient noise to levels that could diminish significantly the range of effective com- munication in soni ferous fish es [ 16 ]. With the auditory brain response (ABR) technique, neural activity is monitored electrically in response to a series of specified exposures to sounds of unique single frequency, duration, and amplitude in order to construct a hearing response curve. Our active echo cancellation procedure makes ABR experiments more readily doable outside of an anechoic environment. It therefore permits a wider range of organisms to be investigated, including aquatic organisms that are difficult (or impossible) to transport to shore side tanks, maintain in captivity, or both. 2. THEORETICAL CONSTRUCT Every chamber or aquarium used for hearing studies has a unique system response to acoustic perturbations that must be characterized prior to generating a stimulus that will be perceived at a particular location as echo-free. Sounds emitted by the stimulus source will reflect from the walls and floor. Thus, at a particular location in the chamber, the sound heard will consist of the sum of the emitted waveform plus reflections. Reflections that involve multiple surfaces are sometimes termed reverb er a - tio n, but for the purposes of this paper (and in the litera- ture, generally), these reflected waveforms are called “echoes”. Echo-cancellation was invented at Bell Labs in the 1960’s. In its most basic form, a time-delayed sample of the emitted signal is subtracted from the received signal. The Least Mean Squares [17] and the Normalized Least Mean Squares algorithms that followed are now imple- mented in software or hardware in a host of modern devices including faxes and telephones to reduce echoes generating by reflections from impedance mismatching in the communications channel, and other effects. The net result of echo-cancellation is an improvement in the signal-to-noise ratio of the received signal. More recen- tly, nonlinear echo cancellation [18] can adaptively can- cel noise in a dynamic environment where the echoes are constantly changing. For example, these algorithms are used in noise-canceling headphones used in noisy envi- ronments. Although there is some evidence that marine mammals like dolphins have neural circuitry able to cancel noise from unwanted reflections from their own emitted sound signals [19], most organisms will perceive the sum of all arriving waveform s with thei r sensory appatus. Most hearing studies are conducted over a range of frequencies; thus it is appropriate to characterize the sys- tem response of the test chamber using a white-noise source. By measuring the system response to white noise at a specific location in the chamber, it is possible to then create an acoustic inverse waveform that nulls reverberations at that location through destructive inter- ference. This “inverse wave” is convolved with a desired stimulus wave (in frequency space), to produce a pre- compensated stimulus waveform. When transformed in- to a time domain signal and played into the chamber, an echo-free stimulus wave will be perceived at the location where the acoustic system response was measured. 3. IMPLEMENTATION USING MATLAB SCRIPTS A set of four interconnected MATLAB scripts: BBTEST.m, MAKESTIMULUSFILE.m, ACOUSTICINV.m, and SE- GMENT.m are used to produce a pre-compensated stim- ulus for location-specific echo cancellation. Each script is described below in chronological order of operation. Copies of the scripts are available for download at: “http://www.vims.edu/about/directory/faculty/brill_rw.php” as “MATLAB scripts.zip”. 3.1. BBTEST.m The BBTEST.m script makes a broadband noise file in signed 16 bit integer format that is then played through the sound emitter to measure the system acoustics (i.e., the acoustic transfer function of the tank and the room). This script must be run prior to running the MAKES- TIMULUSFILE.m script described immediately below. The parameters settable by the user, are: the system fre- quency (in Hz, given by sampersec/2), the lowest and highest frequencies of the “broadband” stimulus to be used for a given set of experiments (lowfreqcps and hifreqcps, respectively, in Hz), the length of the test noise (samlensec, in seconds), and the peak amplitude of the generated signal in volts DC (peakvolts), where –10 to +10 volts DC is assumed to correspond to values of 32,768 to 32,767 in the 16 bit generated noise file, which is named “bb.16”. The BBTEST.m script checks for violations of the Nyquist sampling criterion (sampling
M. R. Patterson et al. / J. Biomedical Science and Engineering 3 (2010) 861-867 Copyright © 2010 SciRes. JBiSE 863 must be at least twice the desired highest frequency for testing) when setting the lowest and highest frequencies, and the system frequency, and advises the user appropri- ately. It also checks that the peak voltage amplitude is set between 0 and 10 volts DC. The generated noise file has all frequencies repre- sented with equal amplitude. The script achieves this by generating a flat spectrum (in frequency space), with frequencies outside the desired window truncated to zero, with zero DC offset. The resulting spectrum is translated back to the time domain using an inverse discrete Fou- rier transform, implemented using a fast Fourier trans- form, and the real part is scaled appropriately by both the peak voltage amplitude, followed by the signed six- teen bit integer format. 3.2. MAKESTIMULUSFILE.m The MAKESTIMULUSFILE.m script writes the acoust- ically pre-compensated stimulu s files. This script creates the following three subdirectories and a parent file direc- tory labeled “director y”: 1) “bbplayfile” — a subdirectory that holds the stochastic differential broadband noise file used to characterize the system acoustics. The BBTEST.m script (see above) is used to produce a broadband noise file in 16 bit signed integer format that is stored in this subdirectory, under the name “bb.16”, 2) “bbrecvfile” — a subdirectory that contains the received signal caused by playing “bbplayfile”. It is equal in length to the file “bbplayfile”, and is stored un- der the name “bb.txt”, 3) “stimulus” — a subdirectory where the generated stimulus file is placed along with two other files: “ones 0.16” and “ones1.16”, which are files filled with -1 and +1 values, respectively. These two files are used to inv- ert the generated sound stimulus signals during presenta- tion to the test subject during an experiment, when the inverted pre-compensated stimulus file is sent to the sound emitter. (Their purpose is to allow mechanical artifacts arising from the sound stimuli themselves to be later removed from the ABR.) The script allows the user to set the following: 1) system frequency (sampersec/2, in Hz), 2) stimulus duration (stimdurm s, in m s), 3) a vector containing the stimulus frequencies to be generated (stimfreqs, in Hz), 4) the phase at the center of the stimulus window (sti- mcentphasedeg, in degrees), 5) the type of window functio n to be app lied to the sti- mulus [Blackman, Hann, Hamming, None, Rectangular (also called “Rectangle”) windows are available], and 6) whether the resulting stimulus file will be acousticall y pre-compensated based on the acoustic characterization of the system contained in the file “bbrecvfile”. 3.3. ACOUSTICINV.m The ACOUSTICINV.m script does not directly interact with the user, unless the transmitted and received broad noise files are not of the same length, in which case the user is notified of the error.The ACOUSTICINV.m scr ipt creates the acoustic inverse response of the system at the location of the test subject. This script opens the file of signed 16 bit integers (“playfile”), the broadband noise file previously played through the sound emitter, as well as a file of equal length that was received at the hydrophone (“recvfile”). The script then reads the data into two signal matrices, and scales the received signal such that its maximum value is equal to that in the pla- yed signal. The discrete Fourier transform of the played signal is then computed using a fast Fourier transform. The resulting transform is subsequently masked (thro- ugh convolution) to deal with the zero es that result from the band-limiting of the played signal. The mask has ones where the discrete transform component was not too small (normalized coefficient > 1/200). A mask is also applied to the received signal to zero out any fre- quency components that were weakly received (normal- ized coefficient < 1/50). The generated inverse signal has a DC component equal to zero, is symmetric in time around the origin, and is inverted. If this pre-compensated signal were played at the same time that the original sound source was played, no sound would be detected at the hydrophone (or microphone). This waveform is thus noise-canceling via destructive interference for the sys- tem at the location of the hydrophone (or microphone). After making an acoustic inverse file using the AC- OUSTICINV.m script, the MAKESTIMULUSFILE.m script checks that the length of the desired stimulus is not longer in time than the acoustic pre-compensation signal used to characterize the system. It then runs the SEGMENT.m script described below. Note that the SEG- MENT.m script is called automatically and uses para- meters previously supplied by the user in the MAKE- STIMULUSFILE.m script, such as the frequency of the stimulus to be generated, the length in ms, etc. 3.4. SEGMENT.m The SEGMENT.m script generates sine wave signals that are windowed by various apodization functions. This script can handle signal lengths that are even or odd in size and produces waves either with no windowing; or with Rectangular, Hann, Hamming, or Blackman win- dowing. It only notif ies the user if an unknown windo w- ing argument is used in the MAKESTIMULUSFILE.m script. The SEGMENT.m script scales the signal appro-
M. R. Patterson et al. / J. Biomedical Science and Engineering 3 (2010) 861-867 Copyright © 2010 SciRes. JBiSE 864 priately for a 16 bit D/A converter with a r ange of -10 to 10 volts DC. After the windowed sine wave is created, its fast Fou- rier transform is computed and convolved with the aco- ustic inverse file. The resulting waveform is inverse Fou- urier transformed. The real part is taken and stored as the pre-compensated stimulus waveform in signed 16 bit int- eger form as a binary file. Finally, the script writes “ones 0.16” and “ones1.1 6” files, which are files filled with -1 and +1 values, respectively, into the same directory as the pre-compensated stimulus files. These files are used to create in-phase and out-of-phase stimulus signals to allow for elimination of electrode movement artifacts caused by the sound stimulus itself through subtraction. To summarize, the scripts must be executed in the following order: 1) Use the BBTEST.m script to produce a broadband noise file in .16 binary format with the user-supplied ba- ndpass, sample rate, sound duration, and sound ampli- tude. The file produced is called “bb.16” and resides in a user-specified directory. 2) Play the “bb.16” file into the test chamber by send- ing the data within file to the sound emitter, taking care to set the D/A output rate to the same sample rate that was used to create “bb.16”. (Note by the Nyquist sampl- ing criterion, the sample rate needs to be twice the high- est frequency to be presented, and preferably at least ten times. For example, a sample rate of 20,000 Hz gives a system maximum frequency of 10,000 Hz, so it would be prudent to outpu t a maximum frequency of 1,000 Hz.) Record the received wave form concurrently from the hydrophone and name it “bb.txt”. Place it in the same directory as “bb.16”. 3) Run the MAKESTIMULUSFILE.m script, specify- ing the directory where the “bb.txt” and “bb.16” are now located, the stimulus duration, the type of windowing, the peak stimulus amplitude, the set of discrete frequen- cies desired in the stimulus files created (each tone will be in its own file), the sample rate to be output, and the phase in degrees at the center of the stimulus. Note that this script also makes two waveforms, each equal in len- gth to the stimulus file, filled eith er with -1’s or 1’s. The stimulus waveform can be multiplied by these files to create waveforms that are 180 degrees out of phase (i.e., of opposite polarity) which can be successively pre- sented to the test subject. This procedure can be used to eliminate electrode movement artifacts caused by the sound stimuli. These files are created solely for conven- ience of the experimenter and are not used in creating the pre-compensated stimulus waveforms. 4. USING THE ACTIVE EHCO CAN- CELLATION TECHNIQUE Our procedures are easily executed by end-users with only modest familiarity with MATLAB, once the system is fully implemented. Alternatively, script execution can be automated, thereby requiring only minimal input from the end-user. The software also allows for the use of Blackman, Hanning, Hamming, and Rectangular win- dow functions to be applied to the stimulus. These win- dow functions have differing effects on the minimum stop-band attenuation (dB) and the width of the transi- tion bandwidth window (Hz). For example the Rectan- gular window has the poorest minimum stopband at- tenuation (21 dB), but the best (smallest) transition bandwidth based on order of filtering, approximately 1.84 (system frequency in Hz) / (order of the filter). By contrast, the Blackman filter has the best attenuation (74dB) and for the same order filter and system fre- quency, a transition bandwidth 605% greater than the Rectangular window (11.13 x system frequency/filter order). (The Rectangular window is also known as a Rectangle window and a Dirichlet window.) All of the included window functions in the code presented are examples of high to moderate resolution windows, with lower dynamic range than some other filters such as the Nuttall or Blackman-Harris windows. This flexibility in windowing the acoustic stimulus allows users to tailor the stimulus to the sensory ability of th e organism under study [20], and has the advantage that both spectral sidelobes and speaker transients are reduced [21]. The effectiveness of the active echo cancellation pro- cedure (using the Tucker Davis Technologies model Sys- tem II described in [13]) is illustrated in Figure 1. In the absence of an anechoic chamber, windowed electronic sinusoidal waveforms delivered to the speaker (A) result in highly distorted acoustic waveforms at the hydroro- phone (B) due to reverberations. In contrast, when pre- compensated electronic waveforms are delivered to the speaker (C), the resultant acoustic waveforms received by the hydrophone (D) are excellent representations of the desired windowed sinusoidal waveforms (A). The sound pressure and particle acceleration thresholds of six sciaenid fishes commonly found in Chesapeake Bay (eastern USA) have been successfully assessed using the ABR and the system described above [13]. Likewise, the system has been used successfully at sea aboard a large oceanographic research ship (the NOAA research vessel “Oscar Elton Sette”) to assess hearing in yellowfin tuna (Thunnus albacares) with the ABR technique (R. Brill, unpublished). Although the steel superstructure of the ship presented a highly echoic environment, the active echo cancellation procedures successfully produced clear sound pulse stimuli required by the AB R technique, allowi ng the
M. R. Patterson et al. / J. Biomedical Science and Engineering 3 (2010) 861-867 Copyright © 2010 SciRes. JBiSE 865 Figure 1. Waveforms generated by the scripts, resulting in echo-cancellation at an omnidirectional hydrophone submerged in a small test tank used to study fish hearing. A = windowed (Blackman filter) sinusoid electronic signal delivered to the speaker. B = windowed non-pre-compensated acoustic signal received at the hydrophone. Note that although the tone burst was set to10 ms, echoes conti nue well past the end of the signa l and the fundamental frequency is corrupted by additive ef fects of reverberations. C = pre-compensated electronic signal delivered to the speaker for echo cancellation (the x-axis is 120 ms, instead of 30 ms, because the source sounds produced long after the tone burst are required to cancel tank reverberations). D. Echo-cancelled acoustic sig nal received at the h ydrophone. delineation of hearing threshold at specific frequencies (Figure 2). 5. LIMITATIONS OF THE ACTIVE ECHO CANCELLATION TECNIQUE There are two limitations to the active echo cancellation technique. The first is th at the juxtaposition of the sound source and all echoic surfaces must remain fixed once the pre-compensated audio files have been generated. Echo- ic surfaces typically include th e water surface (for exper- iments with aquatic organisms in tanks), major hard sur- faces, personnel (if they were near the sound source or test subject when the “bb.16” broadband noise file was played), and the test subject itself. Any changes in the jux- taposition of reflective surfaces will cause the precom- pensated waveform file to fail to produce an echocance- lled signal. Failure of active echo cancellation is easily detected however, if the hydrophone or microphone used t o create the pre-compensated waveforms is left in place and the acoustic stimulus waveforms observed during their presentation to the test subject. Should such deviations occur, the sequence of files described above can be quickly re-run. In principle, the pre-compensated files could be created, and the subject introduced into the ex- per iment al set -up at the s ame pos ition as the hydrophone or microphone. In practice, however, this is impractical because of the likelihood of affecting the juxtaposition of the echoic surfaces. We therefore recommend that compe- nsation be done with subject and hydrophone (or micro- phone) in their final positions. In our experience, this re- quirement is not problematic as subjects generally anesthe- tized, paralyzed, and restrained for ABR studies [13,20]. The second limitation of the pro cedu r e is th at the most effective echo cancellation occurs at the position of the hydrophone or microphone used to receive the original broadband signal. The effectiveness of echo cancellation diminishes with distance from this position (Figure 3). Therefore, the distance between test subject and hydro- phone (or microphone) used to create the pre-compensated signal must be minimized and should preferably be less than 5 cm. The spectral composition of the sound signal Figure 2. Auditory brain response (ABR) recorded from a yellowfin tuna (Thunnus albacares) to a 500 Hz pre-compensated tone burst using a signal pro- cessing system (model System II) from Tucker Davis Technologies (Alachua, Florida, USA), and pro- cedures described in [13]. Experiments were con- ducted at sea, in an interior compartment of a large steel-hulled oceanographic vessel (a highly echoic environment). Sound pressure levels were attenuated in 5 dB steps, and repeated twice at each sound pressure level. Congruence of the ABRs disappears at a sound pressu re of less than 108 dB (re ferenced to 1 µPa at 1 m), indicating that this is minimal detectable sound pressure level.
M. R. Patterson et al. / J. Biomedical Science and Engineering 3 (2010) 861-867 Copyright © 2010 SciRes. JBiSE 866 Figure 3. The degradation of the active-echo acoustic signals recorded in the same situation as those in Figure 1, but with the hydrophone displaced from its original position. These traces demonstrate how the distance between the hydrophone (or microphone) used during creation of the pre-compensated audio files and the test subject can influence the acoustic signal received by the latter. A = hydrophone in the original position when pre-compensation acoustic files were created. B = hydrophone moved 1 cm from its original position. C = hydro- phone moved 5 cm from its original position. D = hydrophone moved 10 cm from its original position. Figure 4. The frequency composition of the active-echo acous- tic signals recorded in the same situation as those in Figure 1, but with the hydrophone displaced from its original position. The frequency composition of the stimulus signal is only mod- erately influenced by distance from the hydrophone (or micro- phone) used to create pre-compensated stimulus files, even up to distances of 10 cm. A = hydrophone in the original position when pre-compensation acoustic files were created. B = hyd- rophone moved 1 cm from its original position. C = hydrophone moved 5 cm from its original position. D = hydrophone moved 10 cm from its original position. is not greatly affected however (Figure 4). The decrease in effectiveness of active echo can cellation with distance from the hydrophone or microphone used during the initial developmen t of the pre-compensated au dio files is an important limitation that must be taken into consid- eration when this procedure is employed. It is possible that with large fishes the acoustic signal received directly by the otolithic organs and signals resulting from vibrations of the swim bladder [22,23] could be different because the signal would be most effectively echo-cancelled at the structure nearest to the hydrophon e used when making the pre-compensated files. Thus our technique, although highly effective at minimizing sig- nal reverberations, should be carefully applied and in- vestigators fully cognizant of its limitations. 6. ACKNOWLEDGEMENTS This work was funded in part by the Pacific Islands and Northeast Fisheries Science Centers (National Marine Fisheries Service, NOAA), and the Pelagic Fisheries Research Program at the University of Ha- waii. AZH was supported by the funds from the Virginia Marine Re- sources Commission and the International Women’s Fishing Associa- tion. MRP thanks J. Svavarsson, University of Iceland, for a place to write during sabbatical. This is Contribution No.3075 of the Virginia Institute of Marine Science, The College of William and Mary. REFERENCES [1] Schuchmann, M., Hübner, M. and Wiegrebe, L. (2006) The absence of spatial echo suppression in echolocating bats. Megaderma lyra and Phyllostomus discolor. Jour- nal of Experimental Biology, 209, 152-157. [2] Parvulescu, A. (1966) Acoustics of small tanks. Marine Bio-acoustics: Proceedings of the Second Symposium on Marine Bio-Acoustics Held at the American Museum of Natural History, New York 13-15 April, 1966, 7-13. [3] Bjørnø, L. and Kjeldgaard, M. (1975) A wide frequency band anechoic water tank. Acustica, 32, 103-109. [4] Darner, C. (1954) An anechoic tank for underwater sou- nd measurements under high hydrostatic pressure. Jour- nal of the Acoustical Society of America, 26, 221-222. [5] Edwards, B. and Nedwell, J.R. (1992) A low frequency underwater anechoic for laboratory applications. Euro- pean Conference on Underwater Acoustics, Elsevier App- lied Science, New York, 60-63. [6] Zhang, J.M. Chang, W., Varadan, V.K. and Varadan, V.V. (2001) Passive underwater acoustic damping using shunted piezoelectric coatings. Smart Materials and St ruc- tures, 10, 414-420. [7] Varadan, V.K., Howarth, T.R., Bao, X.Q. and Varadan, V.V. (1990) Active underwater acoustic damping coating. Journal of the Acoustical Society of America Suppl- ement, 87(S1), S146. [8] Biju, P., Abraham, J.K., Varadan, V.K., Natarajan, V. and Jayakumari, V.G. (2004) Passive underwater acous- tic damping materials with Rho-C rubber–carbon fiber and molecular sieves. Smart Materials and Structures, 13, N99-N104. [9] Prout, J. (1986) Garfield Thomas Water Tunnel (GTWT) flow-through anechoic chamber design, Technical Mem-
M. R. Patterson et al. / J. Biomedical Science and Engineering 3 (2010) 861-867 Copyright © 2010 SciRes. 867 JBiSE orandum, Pennsylvania State University Applied Resear- ch Laboratory, National Technical Information Service, USA. [10] Carlson, T.J. and Popper, A.N. (1997) Using sound to modify fish behavior at power-production and water- control facilities: A workshop, (December 12-13) 1995 Portland State University, Portland, Oregon, Phase II: Final Report, Report to Bonneville Power Administration, Contract No. 1986BP62611, BPA Report DOE/BP- 62611-11, Project No. 19920710. [11] Lovell, J.M., Moate, R.M. Christiansen, L. and Findlay, M.M. (2006) The relationship between body size and evoked potentials from the statocysts of the prawn. Pa- laemon serratus. Journal of Experimental Biology, 209 (13), 2480-2485. [12] Mensinger, A.F. and Deffenbaugh, M. (2000) Anechoic aquarium for ultrasonic neural telemetry. Philosophical Transactions of the Royal Society of London B, 355 (1401), 1305-1308. [13] Horodysky, A.Z., Brill, R.W., Fine, M.L., Musick, J.A. and Latour, R.J. (2008) Acoustic pressure and particle motion thresholds in six sciaenid fishes. Journal of Exp- erimental Biology, 211(9), 1504-1511. [14] Scholik, A.R. and Yan, H.Y. (2001) Effects of underwa- ter noise on auditory sensitivity of cyprinid fish. Hearing Research, 152(1), 17-24. [15] Vasconcelos, R.O., Amorim, M.C.P. and Ladich, F. (2007) Effects of ship noise on the delectability of com- munication signals in the Lusitanian toadfish. Journal of Experimental Biology, 210(12), 2104-2112. [16] Wahlberg, M. and Westerberg, H. (2005) Hearing in fish and their reaction to sounds from offshore wind farms. Marine Ecology Progress Series, 288, 295-309. [17] Widrow, B. and Hoff, M. (1960) Adaptive switching circuits. Western Electronic Show and Convention, Insti- tute of Radio Engineers, 4, 96-104. [18] Kuech, F. and Kellermann, W. (2006) Orthogonalized power filters for nonlinear acoustic echo cancellation. Signal Processing, 86(6), 1168-1181. [19] Nachtigall, P.E., Au, W.W.L., Roitblat, H.L. and Pa- wloski, J.L. (2000). Dolphin biosonar: A model for bi- omimetic sonars. Proceedings of the First International Symposium on Aqua Bio-Mechanisms, 115-121. [20] Kenyon, T.N., Ladich, F. and Yan, H.Y. (1998) A com- parative study of hearing ability in fishes: the auditory brainstem response approach. Journal of Comparative Physiology A, 182, 307-318. [21] Silman, S. and Silverman, C.A. (1991) Auditory diagno- sis: Principles and applications. Academic Press, San Di- ego, California. [22] Yan, H.Y., Fine, M.L., Horn, N.S. and Colón, W.E. (2000) Variability in the role of the gasbladder in fish au- dition. Journal of Comparative Physiology A, 186(5), 435-445. [23] Ramcharitar, J.U., Higgs, D.M., and Popper, A.N. (2006) Audition in sciaenid fishes with different swim bladder- inner ear configurations. Journal of the Acoustical Soci- ety of America, 119, 439-443.
|