Tick 2

Overview

The purpose of this meeting was to review the specification that Eli wrote after some envisioning discussions with Anna:

FlickTok-Functional%20Spec_v0.1_2024May13.docx

Attendees: Greg Wilding Eli Kinney-Lang Anna

Review notes

• FES device
  • It is a “NeuroTrack”, there is no open API, but Ion hacked in a relay so it could be activated by a manual switch / 3.5 mm switch input. This can’t be directly connected to a computer, so another solution will need to be found, ex: a USB triggered relay, etc. This is a risk item.
  • The application will have a manual trigger button in the software so we can test that the connection works. This can be one of the first features we add.
• BCI Essentials will be used for processing of raw EEG from headset, training, producing predictions, etc. The ThinkLab application is an example of how BCI Essentials can be wrapped in a parent application instead of relying on two processes.
• Data capture, aka “saving”
  • EEG data (raw, messages, timestamps, channels)
  • FlickTok messages (timestamp of a swipe, video length before a swipe, comments re: intentionality, trial start/end)
  • There’s probably stuff that was missed / wrong here - need to get into the rest of the application first so we have some common ground.
• Settings
  • Discussion about trials, windows, etc. Will need to get something ~working into Anna’s hands so we can tweak and expose the right settings.
  • A script could be used for the settings instead of a UI (much faster, less user friendly). It would require the user to restart the application each time the settings were changed (maybe not a deal breaker)
  • We could just also have a UI that shows some default settings that the user can change (but they are not persisted).
  • There are no user profiles, this is a single user application
  • Actually, perhaps we do all three here:
    • first - settings are just part of the scripts, ex: settings.py
    • second - create a settings UI to change the important settings on the fly (requires some infrastructure to recreate the EegData object each time). These settings don’t persist.
    • third - add persistence to the settings UI. Make this something that’s later in the project. (however, we still need to design a proper model to share data in the application)
• Training
  • The training mock up shows a red / green box around a video. The idea is that the box would be green when it’s time to for the user to perform the action. 
  • The training will just be automatic instead of letting them press a train button. The user will be stepped through the desired number of trials.
  • There may also need to be an animated action on the side, ex: showing a closing hand (or whatever the action is they’re tying to change).
• Inference
  • In “social media” mode, the red/green box indicates when the system is ready to be used.  There is no signal for “ready”, we’re just going to enforce a ~5-7 second delay before letting the user start.
  • This may be something to put into the settings UI as it may be a per-person setting.
• Why not just use T2S? It can send keystrokes to a web browser to trigger scrolling
  • G.tec headset is the likely target
  • Don’t want to be stuck with Emotiv headset re: access to data, etc
  • Nobody really likes Emotiv
• Why websites? They are harder interact with automatically, can we avoid this pain?
  • Easy access to lots of content and better content is the main reason. This will be better than just dumping some videos into a folder and having the app just cycle through them. Though, worst case, perhaps this is the fall back
  • We could make a simple video player the first step, add in website control later
• What website?
  • Instagram is the first target

Next steps

• Meetings
  • First planning meeting next week (Thursday)
  • Will combine the demo + planning meeting on subsequent Thursdays
• Project management
  • Get project management envrionment set up (Jira project, confluence space, etc)
  • Figure out how to get the simplest app delivered early so we can iterate on settings, etc. (ex: defer settings UI, persistence, animations, website)
  • Fill up backlog
• Plan spikes for risks:
  • How to actually control FES device?
    • existing hardware has a relay switch, but this can’t be connected to the computer. 
    • Need Ion to make something
    • A T2S device could be used, but there are none available, unless one of the very first ones are used.
  • How do we “scroll” to the next video in instagram from an application
    • Maybe just simulate keystrokes?
    • Is there an api for this? (would be specific to each website)
  • How do we get the duration of a video?
    • Not sure, this depends on a website.
    • Discussed why we’re using a website at all, why not a set of videos we can step through.
• Application design spike
  • Single application?
  • Do we need a real browser, or can a QWebViewEngine be used? (we want to be able to put a red/green box around the website - single application makes this easier)
  • MVC design, ie: how to propagate changes to settings
  • Persistence design, ie: how to save values