signal Error Detection for WBSN And data Recovery.

Edwar is a system based on several modules that allows processing of data collected by different devices such as E4 or Everion.

The processing focuses on features extraction, but it can include signal error detection, and signal correction or reconstruction. This core version allows working with electrodermal activity (EDA), interbeat intervals (IBI), accelerometry (ACC) and temperature (TEMP).

The results can be saved as an output variable, CSV files or in a database. To contribute, visit the wiki.

Starting with EDWAR

This core version is based on three modules: Loaders, Parsers and Savers. Loaders read input files from a specific device, so every device has its own Loader. Parsers process these data to calculate features and, finally, results can be saved according to the Saver module employed. User specifies which modules to use in a given configuration (except Saver modules). The following command lets install the configuration file needed.

edwar.install.structure_cfg()

After the installation, the system automatically asks some information, structured in several data fields: device selection, input collection and output collection.

The system can be configurated for several devices, each one with its own files, Loaders and Parsers. Initially there is a default configuration for E4 and Everion devices. If the user wants to edit the configuration for a specific device, this command must be used:

edwar.configure.devices()

The steps followed in installation will be repeated. Finally, user can execute the system with the given configuration:

edwar.Run(device='my_device', path='path_to_data_files').my_saver_module()

There are three different Saver modules:

Saver module	Purpose
get_output()	to save data as aoutput variable
to_csv()	to save data in CSV file
to_database()	to save data in a given database

To get more information about, read Data Saving.

The supermodule Run interconects all modules according the following schema:

Device Configuration

In the first configuration step system requests the device name. According to each device, different data files and variables will be used, so a specific Loader module is required. In this core version, Loaders to read data from E4 and Everion devices can be used. If the Loader needed is not listed, leave it blank. It will be declared as parameter in the run function. In the next steps, variables and features must be declared for this device.

Data Reading

For each device, system asks for the input data, so file names must be given. Every Loader gives facilities to introduce this data, so it is important to know what expects each Loader before introducing the file names. The file path is not relevant, it will be specified in the run function parameters. For each file, the Loader needs the name of the signals in each column. The order is important. In accelerometry file, for example, the variables are x, y, z, not y, x, z or z, y, x. The name of the variables must coincide with the Parser modules input name. Otherwise the Parser will not recognize the signal. These inputs are descibe in Data Processing.

E4

E4 structures the data in several CSV files with the name of the signal as name of file (EDA.csv, IBI.csv...). For the E4 Loader, only the name of the file must be introduced, with no extension (ACC.csv -> ACC).

Everion

Everion file name are long, so the user must introduce only a part, the signal name preferably. For example, for the following IBI file: bop_1533550392847_IBI_5b2cc93e71b0710100a724db_1533160800_1533247259.csv -> IBI

Everion files have several columns with IDs, timestamps, quality of measure... The Loader reads the variable value and timestamp columns. The variable name must be the one expected from the Parser. In case of sweating signal, Everion calls it GSR (Galvanic Skin Response), others use SC (Skin Conductivity), but the Parser module expects a variable called EDA.

Data Processing

There are 4 Parsers. The TEMPparser only returns the input. The ACCparser does not work correctly yet. The following table describes the input variables and the output features calculated in each Parser.

Input Variable	Parser	Output
EDA	EDAparser	EDA, ERROR, SCL, SCR, SNMA
IBI	IBIparser	IBI, ERROR, HR
ACC	ACCparser	HAND, ACT
TEMP	TEMPparser	TEMP

In the last configuration step, all the Parsers to be used must be declared and also the output features to be saved. In case all the features are needed, it is not necessary to list them, only write '*' instead.

Data Saving

To save the output features, there are three posibilities:

As output variable

edwar.Run(device='my_device', path='path_to_data_files').get_output()

The calculated features are presented in a list of DataFrames. Every dataframe has the results from each Parser. Some Parsers generate more than one dataframe. Each dataframe has the Parser name that has generated it (parser_name) and the sampling frequency (frequency). Every dataframe has only the features selected in the configuration.

As CSV file

edwar.Run(device='my_device', path='path_to_data_files').to_csv(path='path_to_output_directory')

Selected output features are saved in several CSV files, one per dataframe. In every CSV file there are a index column with datetimes and one or more feature value columns. As parameter of the Saver module, directory to save data must be specified.

The standard name of the file is "XXXparser_x1_x2_..._xn_date1_date2.csv" where "XXX" is the name of the parser, "x1,x2..., xn" are the variables in the file, "date1" is the time and date when de data was collected and "date2" is the time and date when the file was created. The format fo date1 and date2 are: AAAAMMDDTHHMMSS. (AAAA - year, MM - month, DD - day, T - separator, HH - hour, MM - minute, SS - second). If you want change the name of the file, add csv_name="name_you_want" as parameter of the to_csv method.

In database

edwar.Run(device='my_device', path='path_to_data_files').to_database()

Selected output features are saved in a database. To use this Saver, a database configuration must be done previously. The following command lets install the configuration file needed.

edwar.install.database_cfg()

In this configuration, user introduces database information such as IP, port, user, password... All data are sensible, so this file is encrypted with AES and uses the database password for the given user to decrypt the file (so in every execution of this Saver module, the password is required to save data in desired database).

To change any data, use the following command.

edwar.configure.database()

This module has an optional parameter, data_to_db_function, which is needed when the default function to adapt the output to the database structure does not work properly.