Experimentally parametrising the cocktail party nightmare

Experimental data featured in the publication: Modelling active sensing reveals echo detection even in large groups of bats, Beleyur & Goerlitz 2019, Proceedings of the National Academy of Sciences also see the associated preprint: Active sensing in groups : (what) do bats hear in the sonar cocktail party nightmare?, Beleyur & Goerlitz 2019, BiorXiv preprint, https://www.biorxiv.org/content/10.1101/817734v1 This repository has experimental data collected to : 1) Measure bistatic and monostatic target strengths of a stuffed Myotis myotis bat. 2) Measure the amount of acoustic shadowing multiple bats cause when they fly in groups. Please refer to the Jupyter notebooks in the 'analysis folder' for more details of the analysis and experimental workflows. The zipped folder has *all* audio recordings made from the start to the end of experiments. The contents of each folder, in alphabetical order: analysis_code : Jupyter notebooks that detail the experimental setup and analysis used to process the .WAV files in various other folders. The Jupyter notebooks provide the rationale for an experimental setup and the analysis workflows used. effect_of_lights : Some indoor lights can give off weird electronic and ultrasonic noise. This folder has a few recordings made to check if the lights in the experimental room would affect the recording system. experiment_recordings : series of recordings (raw WAV files) that were used to estimate the monostatic and bistatic target strengths of the stuffed bat, and a sphere (control object). how_much_Atm_abs : A folder that estimated how much of the 'shadowing' I saw (when the mic and speaker were in line with the bat in between) was from atmospheric absorption, and how of the sound level drop was actually shadowing. matched_filter : some trial codes - not used in any way finally > mic_sensitivity : Our lab had many Avisoft CM16 condenser microphones in various working conditions. This folder contains the raw data and analsis Jupyter Notebooks which describe why and how I chose the final microphone used for all proceeding experiments micA02_directionality : Recordings that quantify the directionality of the finally chosen Avisoft Cm16 mic with serial number A02 multi_object_shadowing : Recordings that quantify how much sound is shadowed by the presence of bats themselves. Upto six model bats were hung in a line with microphone and speaker on either side. The amount of shadowing was calculated by the drop in recording level in comparison to no bats. num_instances : HRG had suggested to check the effect of running multiple instances vs one instance of Avisoft USGH recorder software. He had noticed that when recording and playback need to be done they are best done with two instances. Playback + recording worked well on just one instance with my laptop. shadowing : some initial calculations of how much delay would be there between the playback and echo reflecting off the bat at various mic-speaker positions. speaker_directionality : We had a bunch of speaker types in our lab including commercial Avisoft membrane speakers and custom-built Polaroid speakers. I quantified the differences between different models and individual speakers to see which speaker produced the loudest sound. The directionality of each speaker was also quantified. The loudest speaker was then used for all proceeding experiments. The Jupyter notebook and recordings made to determine the loudest speaker and its direcitonality is also in this folder. v_short_pulse : a series of trial recordings to see if the Polaroid speaker was capable of playing back very short pulses (~100 microseconds). Not used to produce any results. video : a series of timelapses made to broadly document the experimental setup and the process of data collection. which_USG416 : we had two Avisoft USG 416's lying around in the lab. I knew that one of them was glitchy and perhaps lost samples while recording. This folder contains a few recordings made to check which one had the lower background noise and if both worked. why_the_blips : at some point during the recordings in Raum 1.03 I began to notice a series of odd spikes in the recordings. The spikes stopped happening after some point. The recordings here were made to try and figure out where the spikes were coming from. Relevant files lying outside in the folder : multi_obstacle_playback_v3.WAV : playback signal used in the acoustic shadowing experiment. The python module used to generate this file is missing - and was likely deleted as an oversight. all_pbks_stereo_1mpbk_version.wav : playback signal used in the bistatic and monostatic target strength experiment. make_playback_sound.py : python module used to create 'all_pbks_stereo_1mpbk_version.wav' Speaker+and+mic+properties.html : An initial report done detailing the speaker and microphone properties. This report was generated before start of experiments. Since there were multiple mics and speakers on hand I wanted to measure them all to understand which particular mic and speaker were suitable for the ensonification experiments. All other files not mentioned specifically by name are a product of iterative testing over the course of the experiment. License All the code and data in this repository are licensed with a Creative Commons Attribution 4.0 International license. Please access the hyperlink for details. Citing this dataset If you use any of the raw data or code in this repository please cite the associated manuscript above, along with the Zenodo DOI of this repository. Further Information For further questions on this dataset please contact me at [e-mail] Last updated : November 2019 , Thejasvi Beleyur

本数据集对应发表于《美国国家科学院院刊》（Proceedings of the National Academy of Sciences）的论文《Modelling active sensing reveals echo detection even in large groups of bats》（作者Beleyur与Goerlitz，2019年），另可参阅相关预印本《Active sensing in groups : (what) do bats hear in the sonar cocktail party nightmare?》（作者Beleyur与Goerlitz，2019年，发布于BioRxiv预印本平台，链接：https://www.biorxiv.org/content/10.1101/817734v1）。

本仓库收录的实验数据用于完成两项研究：1） 测定一只剥制的大足鼠耳蝠（Myotis myotis）的双静态与单静态目标强度；2） 量化集群飞行的多只蝙蝠产生的声学遮蔽效应。有关分析方法与实验流程的详细说明，请参阅"analysis"文件夹内的Jupyter Notebook文件。

本压缩包包含实验全程录制的全部音频文件。各文件夹按字母顺序排序，具体内容如下：
1. "analysis_code"：包含详细记录实验设置与分析流程的Jupyter Notebook，用于处理其他文件夹内的.WAV格式音频文件，本部分同时阐明了实验设计的依据与分析工作流。
2. "effect_of_lights"：部分室内照明设备会产生异常电子噪声与超声噪声，本文件夹收录了用于检测实验室照明是否会干扰录音系统的录制样本。
3. "experiment_recordings"：包含用于估算剥制蝙蝠与球体（对照物体）的单静态、双静态目标强度的原始WAV格式录音序列。
4. "how_much_Atm_abs"：本文件夹用于量化分析——当麦克风与扬声器呈直线排布且蝙蝠位于二者之间时，观测到的"遮蔽效应"有多少来自大气吸收，又有多少源于实际的声学遮蔽。
5. "matched_filter"：部分未在最终分析中使用的试算代码。
6. "mic_sensitivity"：本实验室拥有多台处于不同工作状态的Avisoft CM16型电容麦克风，本文件夹收录了原始数据与分析用Jupyter Notebook，用于说明选择后续实验所用麦克风的依据与流程。
7. "micA02_directionality"：收录用于量化最终选定的序列号为A02的Avisoft CM16型麦克风指向性的录音数据。
8. "multi_object_shadowing"：收录用于量化蝙蝠自身产生的声学遮蔽量的录音数据。实验中将最多6只模型蝙蝠沿直线悬挂，麦克风与扬声器分别置于阵列两端，通过对比无蝙蝠时的录音声级，计算得到遮蔽效应的强度。
9. "num_instances"：HRG曾建议测试Avisoft USGH录音软件运行多实例与单实例的效果差异。其观察到，当需要同时进行录音与播放时，使用两个软件实例效果最佳；而在本实验所用笔记本电脑上，单实例即可实现稳定的播放与录音。
10. "shadowing"：包含初始计算结果，用于分析不同麦克风-扬声器位置下，播放信号经蝙蝠反射后产生的回声与原始播放信号之间的延迟情况。
11. "speaker_directionality"：本实验室拥有多款扬声器，包括商用Avisoft膜片式扬声器与定制的Polaroid扬声器。本部分量化了不同型号与个体扬声器的性能差异，以筛选出声压级最高的扬声器用于后续全部实验。本文件夹同时收录了用于确定最优扬声器及其指向性的Jupyter Notebook与录音数据。
12. "v_short_pulse"：一系列试录样本，用于测试Polaroid扬声器是否能够播放时长约100微秒的极短脉冲信号，该部分数据未用于生成最终研究结果。
13. "video"：一系列延时摄影视频，用于整体记录实验设置与数据采集流程。
14. "which_USG416"：本实验室闲置有两台Avisoft USG 416设备，已知其中一台存在故障，可能会在录音时丢失采样样本。本文件夹收录了测试录音，用于甄别哪台设备的背景噪声更低，以及验证两台设备是否均可正常工作。
15. "why_the_blips"：在Raum 1.03房间进行录音的过程中，曾观测到录音中出现一系列异常尖峰信号，该现象在一段时间后自行消失。本文件夹收录的录音用于排查尖峰信号的来源。

根目录下的相关文件说明如下：
- `multi_obstacle_playback_v3.WAV`：声学遮蔽实验所用的播放信号，用于生成该文件的Python模块已丢失，推测为操作疏漏所致。
- `all_pbks_stereo_1mpbk_version.wav`：双静态与单静态目标强度实验所用的播放信号。
- `make_playback_sound.py`：用于生成`all_pbks_stereo_1mpbk_version.wav`的Python脚本。
- `Speaker+and+mic+properties.html`：实验开始前撰写的初始报告，详细记录了扬声器与麦克风的性能参数。由于本实验可用的麦克风与扬声器型号较多，本报告用于全面测量各类设备，以筛选出适用于声学激励实验的特定设备。
- 其余未单独列明的文件均为实验过程中迭代测试所产生的产物。

许可声明
本仓库内的全部代码与数据均采用知识共享署名4.0国际许可协议（Creative Commons Attribution 4.0 International License）进行授权，详细条款请参阅协议链接。

数据集引用
若您使用本仓库内的原始数据或代码，请引用上述关联论文，同时一并注明本仓库的Zenodo DOI。

更多信息
若您对本数据集有任何疑问，请通过邮箱[e-mail]联系我。

最后更新时间：2019年11月，Thejasvi Beleyur