Add support for Empatica Embrace

[ErraticPattern]

Hello, it would be great if we could get support for Empatica Embrace, a wristwatch wearable widely used for acquiring phyisiological data from PPG, EDA, skin temperature, and acceloremeter sensors.

Context / Motivation

This wearable is widely used in neuroscience research, specially in the context of mobile body-brain imaging, and having support for it would hopefully make BIDS and other tools of physiopy more easily available to the community.

Possible Implementation

Since I already worked with these files I have a good overview on the data types, sampling rates, and data fetching. The raw data is stored in avro files. One avro file is created for each 15 minute data chunk, so the researchers would have to select all the files pertaining to each session. I have code in python to automatically fetch avro files from the AW3 cloud and automatically transform those into timeseries in a tabular format like csv or tsv. The users would simply have to input their credentials. Physiopy could then take the data streams and event streams (all in tabular format) and convert those to be BIDS compatible. I would like to help if necessary, although I am not familiar with the codebase.

Is this feature aligned with physiopy's goals?

[smoia]

Hello @ErraticPattern! Thank you for getting in touch with us.

This is definitely within the scopes of phys2bids and physiopy. In fact, now that we are working on proposing better BIDS physiological specs (to better accommodate data like wereables between the rest), this type of data is what we expect to deal with.

I am not aware with this type of data, so I don't know how far I can help you with understanding which bit is which, however:

• If you want to contribute, you can write a function for the I/O, modifying this file. Consider, however, that there is some data that needs to be present, or you need to find a way to pass it.
• We would need a sample of the data to understand what we're working on and to test the function, and this data should go into our OSF repository. Could you provide some? It doesn't have to be massive, in fact, 10-100 samples should be enough (what we mainly care about is the data structure!).
• I'm more and more inclined to drop the necessity of a trigger channel (you can get away now by setting anything to it). That may require some tweaking of the main workflow (although since we support GE data, we shouldn't need it anymore).
• This week I'll be at a hackathon, and I might work on generalising phys2bids, so there might be a chance to put in some work towards making this issue easier to complete.

Let me know how you want to continue!