Page 24 - Acoustics Today

ance from one of the authors. They represent relatively favor- able scenarios. The great grey owl study had the advantage of a limited number of other sounds that could occur within the frequency bands of interest at night. The vehicle study was able to utilize a generalized template for low resolution spectro- gram correlation to detect a wide range of vehicles.
When standardized measurements are used to character- ize the structure of sounds, several principles should guide the design and implementation of these measurements. Measurements must be relevant to the classification task, and analysis and presentation will be eased if the values are read- ily interpreted, if they summarize salient physical or spectro- graphic features. In order to minimize “the curse of dimen- sionality”23 and unintentional weighting of particular charac- teristics, measurements should be at least partially uncorre- lated across the sample. Lastly, measurements should be robust, yielding similar values when there are modest changes in the acoustical background behind the signal of interest. It may be fruitful to consider measurements as extremely lossy data compression, to evaluate the relative merits of alternative sets of measurements in terms of the accuracy of signal reconstruction.
Perhaps the most challenging aspect of automated detec- tion is quantifying the fraction of events that the detection and classification regime failed to capture. The straightforward practice is extensive review of a randomized sample of the raw data, but this will be laborious and inefficient when signals of interest are rare. Another option is to utilize multiple methods in conjunction with an approach analogous to multisensory data fusion24 or mark-recapture analysis25 to estimate the frac- tion of signals that have been missed by all methods.
Conclusion
Autonomous acoustical monitoring is an emerging tool for terrestrial studies of ecology and animal behavior, and compelling results will be forthcoming as systems improve and researchers become familiar with their features and idio- syncrasies. In terms of data collection, there are numerous commercial options for microphones and recorders, orders of magnitude differences in performance, and more than an order of magnitude difference in price. Brent Hetzler of the National Park Service26 has conducted extensive acoustical
surveys for Mexican spotted owls using $20 voice recorders that he modified for long-term operation. Digital audio recorders capable of broadband, high fidelity recording gen- erally cost $200 or more, and specialized acoustical monitor- ing equipment can cost $2000 or more.
The right recorder will rarely be the most expensive unit, and it may not be the unit with the highest performance/price ratio. Spreading and absorptive losses combine with refraction in atmospheric sound propagation to present severe limits to the maximum range of detection. Distributed sensing allows for effectively higher signal to noise ratios through an abundance of vantage points, espe- cially when the location of the source of interest is unknown. In addition, acoustical sampling may need to span a wide range of environmental conditions and a single sensor is less robust to component failure. Accordingly, many units that are less capable may yield more useful information than a few
27
ters of interest—occupancy, temporal activity patterns, spa- tial patterns of movement, population density, breeding activity, and possibly—when long-term individual recogni- tion is possible—individual survivorship and reproductive output. The extent and nature of the necessary acoustical data, as well as supplemental information, depends upon the parameter of interest and the scope of the study. Useful infor- mation will undoubtedly emerge from any study that collects good recordings, but efficiency and effectiveness will be greatly improved by developing an explicit sampling plan that addresses one or more parameters for estimation. Focused monitoring effort will yield more decisive results
28
units with more impressive performance.
Bioacoustical monitoring can document many parame-
with less effort than undirected effort.
References
AT
1 R. Bradley, “Making animal sound recordings,” Am. Birds 31, 279–285 (1977).
2 J. Sueur, S. Pavoine, O. Hamerlynck, and S. Duvail, “Rapid acoustic survey for biodiversity appraisal,” PLoS One 3, 1–9 (2008).
3 P. Newman and R. Manning, and K. Treviño, “From landscapes to soundscapes: Introduction to the special issue,” Park Sci. 26, 2–5 (2010).
4 D. T. Blumstein, D. J. Mennill, P. Clemins, L. Girod, K. Yao, G.
OFF
ON
AqFlex : Revolutionary on/off absorption system
• Lowers RT (63 - 1000 Hz) by app. 45% at the push of a button
• αon = app. 0.5 (63 - 1000 Hz) in entire audience/stage area. αoff = app. 0.0
• For use in music education institutions , performing arts centres, etc
• Complies with the ASTM E 84, NFPA 701 standards
• Price: app. 18-20 $/ft 2 ceiling area
• First US-installation: half price. Then US importer will be appointed.
• US Patent # 7905323
Any kind of acoustics at any time
Bioacoustical Monitoring in Terrestrial Environments 23
flexac.com