MEMS-cochlea: Dataset for publication

This is the dataset to the publication " Neuromorphic acoustic sensing using an adaptive microelectromechanical cochlea with integrated feedback" by Lenk et al. (DOI will follow soon). Explanation of data: 1.) 'MEMS cochlea response to natural sound' (dataset for fig 2 in publication): In this dataset, the file "natural sound dateset" was used to drive a loudspeaker. Its given in wav-format. File named timeseries..." give the data of the response of two different sensors as well as a measurement microphone (named "input") to the wav-file. First column time in sec, second column sensor signal in V. Files named "powerspectra..." give the power spectra data of the three time series, first column frequency in Hz, second column power in absolute values not dB. 2.) 'Sensor response in dependence of feedback' (dataset for fig 3 in publication): The dataset includes the sensor signal amplitudes in mV (2nd column) as function of sound pressure amplitudes in Pa (1st column) in files with name starting "fig3a..." for different feedback strengths a_f given by the filename. Files, whose names start with "fig3b+c", give the gain (sensor amplitude active, i.e. a_f>0, divided by sensor amplitude passive, i.e. a_f=0) in the 2nd column as a function of the feedback strength a_f (1st column). In files named "fig3e_sensamp...", the sensor signal amplitude in V (2nd column) is given in dependence of the feedback strength a_f (1st column) for different driving voltages of the loudpseaker, given by the number after "loud" in the filename. If the filename says "negafnegDC", the feedback strength a_f is negativ. If it says "posafnegDC", the feedback strength was positive. The DC voltage of the feedback was always -200mV. From the dependence of sensor signal amplitude on the driving signal amplitude (both in mV), the sensitivity is extracted as the slope of the curve in mV/mV. The sensitvity is given in files, named "fig3e_sensitvity..." in the 2nd column as function of feedback strength a_f (first column). 3.) 'Comparison experiment vs model' (dataset for fig 4 in publication): These files give the values plotted in the graphs. The files, named "acrit..." contain the values of a_crit (feedback strength at bifurcation, 2nd column of file) as function of bias voltage u_DC in mV (first column of file) obtained either from experiment or from the formula (last equation in methods part). The file, named "sensitivity...", has the feedback strength a_f in the 1st column and the sensitivity in nm/Pa, obtained frome xperiments, in the 2nd column. The files, named "effective_Q_factor...", have the feedback strength a_f in 1st column and the effective Q factor, obtained from simulations, in the 2nd column. 4.) 'Two coupled sensors' (dataset for fig 5 in publication): The files contain the frequency response, i.e. power spectral density in dB (2nd column) as function of frequency in kHz (1st column), of two different sensors for different values of the coupling strength, given by the value after "b" in the filename. 5.) 'Dynamic_adaptation_with_code' (dataset for fig 6 in publication): This dataset contains files, names starting with "timeseries...", which give the time series (sensor signal in mV vs. time in sec) for two different driving voltages of the loudspeaker (given by the value after "sound" in the filename), which are shown in fig. 6b in the publication. Files, named "envelope...", give the extracted envelope of the modelled sensor signal in V (2nd column) as function of time in sec (1st column) for different modelled sound inputs, as shown in fig 6c. The envelope was extracted with the program "env.m", written in Matlab. The program "spice_sim" is used to start the LTSpice simulations for adaptation with different parameters. The files in the zip-archive "adapt_,8_,5_natelec" incorporates the necessary files for the LTSPice simulation of the adaptation process.

本数据集对应Lenk等人发表的论文《Neuromorphic acoustic sensing using an adaptive microelectromechanical cochlea with integrated feedback》（DOI将后续公布）。数据说明如下：
1.) "微机电系统（Microelectromechanical Systems, MEMS）耳蜗对自然声的响应（对应论文中图2的数据集）"：本数据集使用"自然声数据集"（原文拼写为natural sound dateset，疑为笔误）文件驱动扬声器，该文件为WAV格式。名为`timeseries...`的文件包含两路不同传感器以及测量传声器（命名为"input"）对该WAV文件的响应数据：第一列为时间（单位：秒），第二列为传感器信号（单位：伏特）。名为`powerspectra...`的文件包含上述三路时间序列的功率谱数据：第一列为频率（单位：赫兹），第二列为功率绝对值（非分贝形式）。
2.) "反馈依赖的传感器响应（对应论文中图3的数据集）"：该数据集包含以声压幅值（单位：帕斯卡，第一列）为自变量、传感器信号幅值（单位：毫伏，第二列）为因变量的数据，对应文件名以`fig3a...`开头的文件，其中不同反馈强度`a_f`可通过文件名获取。文件名以`fig3b+c`开头的文件，其第二列为增益值（即反馈激活时`a_f>0`的传感器幅值除以反馈关闭时`a_f=0`的传感器幅值），以反馈强度`a_f`（第一列）为自变量。名为`fig3e_sensamp...`的文件中，第二列为传感器信号幅值（单位：伏特），以反馈强度`a_f`（第一列）为自变量，对应扬声器的不同驱动电压可通过文件名中"loud"后的数值获取。若文件名包含"negafnegDC"，则反馈强度`a_f`为负值；若包含"posafnegDC"，则反馈强度为正值。反馈的直流电压始终为-200mV。通过传感器信号幅值与驱动信号幅值（二者单位均为毫伏）的依赖关系，可提取得到灵敏度，即曲线斜率，单位为mV/mV。名为`fig3e_sensitivity...`（原文拼写为sensitvity，疑为笔误）的文件中，第二列为灵敏度，以反馈强度`a_f`（第一列）为自变量。
3.) "实验与模型对比（对应论文中图4的数据集）"：此类文件包含绘图所用的数值。名为`acrit...`的文件包含分岔点处的反馈强度`a_crit`（第二列）随偏置电压`u_DC`（单位：毫伏，第一列）变化的数据，该数据分别通过实验与方法部分最后一个方程对应的公式计算得到。名为`sensitivity...`的文件中，第一列为反馈强度`a_f`，第二列为通过实验得到的灵敏度（单位：nm/Pa）。名为`effective_Q_factor...`的文件中，第一列为反馈强度`a_f`，第二列为通过仿真得到的有效品质因数。
4.) "双耦合传感器（对应论文中图5的数据集）"：此类文件包含两路不同传感器的频率响应数据，即以频率（单位：千赫兹，第一列）为自变量、功率谱密度（单位：分贝，第二列）为因变量，耦合强度可通过文件名中"b"后的数值获取。
5.) "基于代码的动态自适应（对应论文中图6的数据集）"：本数据集包含文件名以`timeseries...`开头的文件，其中记录了扬声器不同驱动电压下的时间序列数据（传感器信号幅值，单位：毫伏，随时间（单位：秒）变化），对应论文图6b的内容，扬声器驱动电压可通过文件名中"sound"后的数值获取。名为`envelope...`的文件包含提取得到的建模传感器信号包络（单位：伏特，第二列）随时间（单位：秒，第一列）变化的数据，对应不同建模声输入的结果，即论文图6c的内容。该包络通过MATLAB编写的程序`env.m`提取得到。程序`spice_sim`用于启动针对自适应过程的LTSpice仿真，可配置不同参数。压缩归档文件`adapt_,8_,5_natelec.zip`包含了自适应过程LTSpice仿真所需的全部文件。