LSL Raw EEG Values Interpretation

[Profex04]

Hi,

Have been using pylsl with an Emotiv EEG headset, and facing an issue comparing EDF eeg recordings with online pylsl raw data. Emotiv scales the recorded data to microvolt range when saving the recordings as EDF files, while raw pylsl data pulled from inlet seems to be in an arbitrary dynamic range units in the 4100s-4200s. What would be a sensible scaling factor to compare two datasets meaningfully, there also seems to be an offset as pylsl data is always positive? Went through both Emotiv LSL and LSL documentations, any ideas or help is appreciated.

[hoangphamemotiv]

Hi @Profex04 ,

When you live stream EEG data from EmotivPRO via the Lab Streaming Layer (using pylsl), the data you receive is already in microvolt (µV). This is the same data that gets recorded into the EDF file through the recording feature, there is no difference in terms of signal content.

The EEG raw values you are seeing (~4200 range) are expected. They include a DC offset, typically around ~4201. So the actual EEG signal is the small variation around this baseline.

Note that depending on the headset type, this offset may vary or may not be present. Could you please confirm which Emotiv headset you are using?

In the EDF file, we store this as physical values, along with the corresponding digital scaling factors. So both LSL stream data and EDF recordings should be consistent.

If you are seeing EDF values outside of this expected range, please share a sample file so we can take a closer look.

[Profex04]

Hi @hoangphamemotiv

Thank you so much for the fast response! I have 2 different Emotiv headsets, both Insight and EPOC X. I am seeing these large LSL values for the both.
As for the EDF recordings, they seem to be in the expected microvolt range. I was only confused by the 4100-4200 values for LSL instream, for the both headsets that I have.
Now then, shall I just substract an offset value 4201 etc from the LSL values? How can I find out the specific offset value for each device?

Thank you.

[hoangphamemotiv]

Hi @Profex04,
The ~4100–4200 values in the LSL stream are just a DC offset (baseline) added by the device. The actual EEG signal is the small fluctuation around that value. In practice, you don’t need to worry about the exact offset value or manually find something like “4201” for each device.

For most EEG analysis, we typically:

• Subtract the mean per channel, or
• Apply a simple high-pass filter, or
• Use baseline correction (for ERP)
  All of these effectively remove the offset.

Important point: The offset itself does not carry useful EEG information, and standard analysis methods (PSD, band power, ERP, etc.) either ignore or remove it automatically.

So yes, if you want to visualize or preprocess the data, you can simply do something like: subtract the channel mean → that’s already sufficient. No need to worry about device-specific offset values