A system for acoustic detection, classification, and localization of terrestrial animals in remote locations
A system for acoustic detection, classification, and localization of terrestrial animals in remote locations
复制标题
DOI：
10.1121/1.4970003
复制
发表时间：
2016-11
影响因子：
2.4
通讯作者：
D. Hawthorne;W. Horn;D. Reinke
中科院分区：
文献类型：
--
作者：
D. Hawthorne;W. Horn;D. Reinke
对濒危陆生动物的声学监测通常涉及将观察设备部署到人类观察者难以或难以接近的偏远地点。我们提出了一个系统，旨在部署到这样的位置，这是能够检测，分类和本地化范围广泛的发声。电池/太阳能供电系统可以连续两周记录数据，同时发送实时Wi-Fi检测警报。通常，部署三个系统，每个系统由传感器阵列组成，传感器阵列由16元件麦克风阵列、辅助处理器盒以及所需的辅助设备组成，从而允许检测地理定位。计算机系统使用计算效率高的波束形成算法来实现12 dB的增益并定位方位角和俯仰角的检测。在声谱图图像域中使用模板匹配算法对发声进行分类。定位是通过交叉从三个波束形成器获得的远场方位来实现的。
Acoustic monitoring of endangered terrestrial animals often involves deploying observation equipment to remote locations which may be difficult or hazardous for human observers to access. We present a system designed for deployment to such locations which is capable of detecting, classifying, and localizing a wide range of vocalizations. The battery/solar powered system can continuously record data for two weeks while sending real-time Wi-Fi detection alerts. Normally, three systems, each composed of a sensor array consisting of a 16 element microphone array, an ancillary processor box, plus ancillary equipment as needed, are deployed allowing detection geolocation. The computer system uses a computationally efficient beamforming algorithm to achieve 12 dB of gain and locate detections in both azimuth and elevation. Vocalizations are classified using a template matching algorithm in the spectrogram image domain. Localization is achieved by crossing the far-field bearings obtained from the three beamformer...