Bridging the gap between human expertise and artificial intelligence - ZFMK’s contribution to making the AmiBio concept work

Please download to get full document.

View again

All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
 4
 
 

Sales

  Bridging the gap between human expertise and artificial intelligence - ZFMK’s contribution to making the AmiBio concept work
Share
Transcript
  See discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/258450971 Bridging the gap between human expertise andartificial intelligence - ZFMK’s contribution tomaking the AmiBio concept work   ARTICLE  · APRIL 2011 READS 30 6 AUTHORS , INCLUDING:Olaf JahnZoological Research Museum Alexander Ko… 151   PUBLICATIONS   98   CITATIONS   SEE PROFILE Klaus RiedeUniversity of Bonn 128   PUBLICATIONS   388   CITATIONS   SEE PROFILE Renate ElzenResearch Museum Alexander Koenig 25   PUBLICATIONS   38   CITATIONS   SEE PROFILE Theodoros KostoulasUniversity of Geneva 29   PUBLICATIONS   93   CITATIONS   SEE PROFILE Available from: Olaf JahnRetrieved on: 27 October 2015    2  A MI  B IO   B RIDGING   THE  G AP   BETWEEN  H UMAN  E XPERTISE   AND  A RTIFICIAL  I NTELLIGENCE   ZFMK’ S  C ONTRIBUTION   TO  M AKING   THE  A MI B IO  C ONCEPT  W  ORK   The provision of biological baseline data is one of the main responsibilities of the Zoological Research Museum A. Koenig (ZFMK) in the AmiBio project. Species have to be inventoried and recordings of their vocalizations have to be compiled, identified, tagged, and archived in order to calibrate the sound identification software (AmiBio Newsletter, July 2010 issue, www.amibio-project.eu/en/resources/dissemination/newsletter/2010-july  ). Without this preparative work the automatic acoustic monitoring of wild animal species at Hymettus would not be possible. When the biodiver-sity assessment officially finished in July 2010, we were well aware of the fact that our lists of sound-producing insects, amphibians, reptiles, birds, and mammals were not entirely conclusive. For instance, apart of the 116 bird species that already were known to occur in the project area, we estimated that 150+ additional species might occur there at least seasonally on migration. Therefore, the first author continued to collect field data in December 2010 and in February, March, and April of the present year. During the same period we also continued the sound recording campaign at five independent locations, using commercially available Song Meter recorders of Wildlife Acoustics Inc. ( www.wildlifeacoustics.com/sm2_platform.php ). Until the end of April we gathered over 2000 hours of stereo re-cordings (48kHz, 16 bit), corresponding to roughly 1.5 TB of audio data. During the next several months these re-cordings will be used for extracting additional reference vocalizations and for testing the species-specific detectors with real Hymettus recordings. On the basis of the above-mentioned observer-based bird surveys, the Hymettus bird list has grown to 122 species and the number of species that were recorded by ZFMK staff from 68 to 88 species, corresponding to 72% of the total. At least nine bird species were not previously reported in the literature on the avifauna of Hymettus. From the latter, all but one were already on our list of potentially occurring species. The species not included in the aforementioned list was a big surprise indeed: it refers to the first record of the Lemon-rumped Warbler Phylloscopus proregulus  for Greece. On 5 April 2011, in the park-like area that belongs to the Monastery Kaisariani, the first author of this article took notice of a rather complex warbling song, which he could not identify immediately. While searching for the bird he discovered a very small warbler with a whitish underside, broad yellowish-white superciliary, and two distinct whitish wing bars (see cover of this newsletter). Although he could neither see the distinctive coronal strip nor the yellow rump of the bird, because it moved around in the crown of a pine about 12 meters above the ground, a positive identification was still possible by comparing the song heard in the field with vocalizations of the other look- alike leaf ales at www.xeno-canto.org . Subsequent searches in the Internet revealed that the bird was previously discovered and photo-graphed by Michalis Kotsakis and by several other observers on 2 April 2011 and the following days (see photos at http://deepinbirds.wordpress.com/2011/04/04/phylloscopus-proregulus-update , www.facebook.com/photo.php?   fbid=10150187718288933&set=o.175254428030&type=1 ,  and www.ppetrou.gr/2011/04/444 ). The Lemon-rumped Wale o Pallas’s Wale, as it is soeties alled afte its desie eeds i southe “ieia, othe Mogo- lia, and northeastern China. It is a long-distance migrant, which mainly winters in subtropical southern China and north- eastern Indochina. However, a tiny fraction of its population winters in western Europe, which explains the unusual spring record in Greece. Less eitig, ut uh oe ipotat fo the outoes of the AiBio Pojet, is )FMK’s ok o the efeee li- brary of animal sounds for the Hymettus. During the past months we have been compiling 11,000+ sound recordings for the 175 sound-producing animal species already recorded there as well as for the 148 potentially occurring bird species. The sources for this comprehensive sound collection were online archives, such as Xenocanto ( www.xeno-canto.org ) , McCaulay Library ( http://macaulaylibrary.org/index.do ), SysTax ( www.biologie.uni-ulm.de/systax/portal/ index.html  ), and the Tierstimmenarchiv ( www.tierstimmen.org ). In addition we extracted hundreds of recordings from commercial audio publications. However, compiling the sounds is not enough for making automatic computer-based biodiversity monitoring feasible. Evidently, mathematical algorithms for sound classification are not as flexible as hu-man brains in learning animal sounds. After a few test runs with a tiny subsample of the aforementioned audio ( continuation on Page 3 ) by O. Jahn, K. Riede, V. Eschen, K.-L. Schuchmann, R. van den Elzen, and T. Kostoulas Project progress  3  A MI  B IO   Project progress archive, it was clear that the recordings were not clean enough for developing species-specific sound detectors. The reason is that most audio recordings were made in natural environments and do not only contain the target species but also many other kinds of natural and anthropogenic audio signals. Although some types of noises, such as wind, rain, and airplanes, can be removed by filtering, current sound classification software cannot successfully learn the vocaliza-tions of a target species unless the reference recordings have a good noise-to-signal ratio. That is, they must be virtually free of undesired background species and other noises. Therefore we started to manually select those sections of the recordings that contained exclusively call notes and song phrases of the target species. This is a laborious endeavor in- deed! We estiate that aout 70% of the oigial taget speies’ oalizatios ust e disaded duig the poedue because they overlapped with other (mostly animal-produced) sounds in time and frequency range. In order to comply with the conservation-oriented aims of the Life+ program we are currently focusing on those species that are protected under EU legislation (examples of such species have been mentioned in AmiBio Newsletter, October 2010 issue, www.amibio-project.eu/en/resources/dissemination/newsletter/2010-october  ). In other words, the sound detectors will be developed first for the conservation-dependent species and then for more common ones. However, for some rarer species we will have to add sound examples of our own passively collected Song Meter recordings (see first paragraph) in order to obtain enough examples for training the detectors. All in all this task alone will keep us busy until the end of the AmiBio project in July 2013.  LIFE+ N   ATURE     AND  B IODIVERSITY    4th Issue April 2011 CC ONTENTSONTENTS    Pages 2 & 3 Bridging the gap between human expertise and artificial intelligence Public awareness on Hymettus area conser-vation plans  Page 4  AmiBio meets OpenUp!, the new link between Europeana and biodiversity data resources Contact Us Nikos Fakotakis, Project Coordinator   Wire Communication Laboratory, University of Patras, 26500 Rion-Patras, Greece E-mail: fakotaki@upatras.gr Phone: +30 2610 996 496 http://www.amibio-project.eu/ WWW  .  AMIBIO - PROJECT  . EU    LIFE08 NAT/GR/000539   With the contribution of the LIFE financial instrument of the European Community
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x