Accessing E4 data: The only way to access data from the E4 is via Empatica's Web Portal, E4 connect. The interface allows researchers to visualize E4 session data, download sessions in CSV format, and flag them for permanent deletion.
Study management features: There are study management features that allow for bulk download of data from E4 connect. The study management features are currently in beta - please ask the Empatica support team if you want to use them with your Connect account. See this Empatica Connect documentation PDF for details.
Accessing E4 connect via API: E4 data hosted on Connect can only be accessed through the graphical user interface (GUI) of the web-app. There are no E4 connect APIs for customer use at this time.
File Size: E4 data varies by activity level and other factors. The average file size of a compressed session is just under 1 MB/hour while an uncompressed session is typically just under 3 MB/hour.
Limits to Storage: There are no storage limits with E4 data on E4 connect. Your purchase entitles you to the storage of raw data for the life of your E4 wristband. Please note that Empatica may implement paid storage options for other product lines.
E4 connect Export Formats
E4 connect allows you to access raw data in CSV format for all of your Empatica E4 data.
Data can be downloaded as a compressed directory (ZIP) containing the following files
- ACC.csv - Data from 3-axis acceleometer sensor in the range [-2g, 2g]. (sampled at 32 Hz)
- BVP.csv - Data from photoplethysmograph (PPG). (sampled at 64 Hz)
- EDA.csv - Data from the electrodermal activity sensor in μS. (sampled at 4 Hz)
- IBI.csv - Inter beat intervals. (intermittent output with 1/64 second resolution)
- TEMP.csv - Data from temperature sensor expressed in degrees on the Celsius (°C) scale (sampled at 4 Hz)
- HR.csv - This file contains the average heart rate values, computed in spans of 10 seconds
More details here: E4 data - HR.csv explanation - info.txt - Descriptions of the files
Each sensor's sample rate is hard coded in firmware and optimized to capture the frequency content of relevant signals.
The format of each of these files is described below:
IBI.csv
1381290418.000000, IBI <-- t0, initial time of the data (unix timestamp or seconds from 1-1-1970 in UTC) + name
7.734375,0.875000 <-- sequence of IBIs : t, IBI(t) where t are the seconds from t0
8.609375,0.953125
for a sequence of correctly detected IBIs you have
t1 , ibi(t1)
t2 , ibi(t2)
t3 , ..
and t2 - t1 = ibi(t1)
*Note that IBI data is reported only when BVP signal is clear. Gaps in the IBI data are identifiable when t2-t1 ≠ ibi(t2).
ACC.csv
1380399004.000000, 1380399004.000000, 1380399004.000000 <--- t0, initial time of the data for all channels
32.000000, 32.000000, 32.000000 <-- sampling rate for all channels
27,-25,50 <-- x, y, z value of acceleration at sample #1**
27,-24,50 <-- x, y, z value of acceleration at sample #2, etc.
27,-24,51
26,-23,51
**For analytic purposes you can interpret the raw data as follows:
- xg = x * 2/128
- a value of x = 64 is in practice 1g
BVP.csv , EDA.csv, TEMP.csv
1380399004.000000 <--- t0, initial time of the data
64.000000 <-- sampling rate
-0.251772 <-- value
-1.208777
-2.865649
-1.993523
-0.273807
0.535328
If you haven't found the answer to your question, click the button below and complete the typeform. We will receive all the information needed to assist you further.