%0 Conference Proceedings %T Odontocete click train deinterleaving using a single hydrophone and rhythm analysis %+ AP %+ Lab-STICC_EB_CID_TOMS ; AP %+ GIPSA - Signal Images Physique (GIPSA-SIGMAPHY) %A Le Bot, Olivier %A Bonnel, Julien %A Mars, Jerome, I. %A Gervaise, Cedric %< avec comité de lecture %( ICA 2013 proceedings - Volume 19 of POMA (Proceedings of Meetings on Acoustics) %( ICA 2013 proceedings - Volume 19 of POMA (Proceedings of Meetings on Acoustics) %B ICA 2013 - Acoustics 2013 - 21st International Congress on Acoustics - 165th Meeting of the Acoustical Society of America %C Montreal, Canada %V 19 %P xx %8 2013-06-02 %D 2013 %R 10.1121/1.4800655 %K Acoustical detection of marine life %K passive and active %K Sound production by animals: mechanisms %K characteristics %K populations %K biosonar %Z Engineering Sciences [physics]/Acoustics [physics.class-ph] %Z Physics [physics]/Mechanics [physics]/Acoustics [physics.class-ph]Conference papers %X Most odontocetes live in pods of several individuals, resulting in an overlapping of click trains recorded by passive acoustic monitoring systems. Localization algorithms and click classifiers are usually used for train separation. However, their performances fall down if individuals are too close to each other or if acoustical parameters vary greatly from click to click, respectively. Assuming odontocete clicks follow rhythmic patterns, we propose to use a rhythm analysis to separate mixed click trains from a single hydrophone. The proposed algorithm is based only on inter-click-intervals (ICI) to cluster clicks into trains. It uses information given by complex-valued autocorrelation to compute a histogram which will exhibit peaks at ICIs corresponding to interleaved trains. By this technique, subharmonics corresponding to multiples of ICIs are automatically suppressed. The algorithm is then extended by a time-period analysis leading to a time-varying ICI spectrum. A threshold can be applied on this spectrum to detect the different interleaved trains. The final result is a binary time-ICI map on which trains can be fully and easily distinguished and extracted. We validate it on simulated and experimental data and we show that the algorithm is particularly suitable as a preprocessing tool prior to localization and classification schemes. %G English %L hal-00843208 %U https://hal.science/hal-00843208 %~ UNIV-BREST %~ INSTITUT-TELECOM %~ ENSTA-BRETAGNE %~ UGA %~ CNRS %~ UNIV-GRENOBLE1 %~ UNIV-PMF_GRENOBLE %~ UNIV-GRENOBLE3 %~ UNIV-UBS %~ INPG %~ OSUG %~ GIPSA %~ GIPSA-DIS %~ GIPSA-SIGMAPHY %~ ENSTA-BRETAGNE-STIC %~ ENIB %~ LAB-STICC_ENIB %~ LAB-STICC %~ UGA-TEST-BIS %~ UGA-TEST-TER %~ UGA-TEST-QUATER %~ INSTITUTS-TELECOM