qutang/riars

riars's Issues

Annotate the time of "switching" automatically

Use "switch" voice instead of "beep" as a reminder

Verify laptop can handle 3 sensors at 100 Hz + bluetooth earphone

Estimate and record inference delay

We may use the first window to estimate the inference delay given that we know the time spent for data collection. This estimation ignores the delay in data transferring.

Display the estimated inference delay in the monitor interface.

Beep voice feedback for activity switching

The system won't know if prediction is correct or not, annotator has to provide feedback.

If the prediction is correct for three consecutive windows, beep to switch on the start of the next window.
If the prediction is wrong or uncertain, count until 1-2 minutes (8-10 consecutive windows) then beep to switch on the start of the next window.

Add timer for current variation (in addition to current annotation)

Make changes to voice feedback

1. Use "Keep going" and "Do it differently" to replace voice feedback and switches

Add extra annotation to flag out-of-scope movement

If the variation is beyond the general definition of the ontology, flag it as out-of-scope.

Make each React component testable

Optimize Prediction Monitor

Use first prediction to calibrate the start and stop time of each data window, and the latency for model computation.

Display "Calibrating timestamp" upon receiving the first prediction.

1. Calibrate data window time
2. Calibrate model computation time

Use second prediction to get the participant ready

Display “Getting ready...", and start a five-second count down when there are five seconds left.

Start from the third prediction, make prediction feedback

1. Display "Collecting data for n seconds..." in the first column panel
2. Display "Making prediction..." in the second column panel
3. Change "Making prediction..." to the actual prediction in the second panel

Change voice feedback interval range

With shortened voice feedback text, we may be able to try voice feedback interval as short as one unit of (window size + inference delay) and may extend it to multiple units of (window size + inference delay).

Depending on hardware and number of sensors, the inference delay can increase.

Device	Number of sensors	Inference Delay (estimated sec.)
Macbook Pro i7 16g RAM	1	?
Macbook Pro i7 16g RAM	2	?
Macbook Pro i7 16g RAM	3	5-6
Dell Workstation e5 64g RAM	1	4>2
Dell Workstation e5 64g RAM	2	?
Dell Workstation e5 64g RAM	3	?

Add another "variation" button

"Variation I" and "Variation II", these two should be mutually exclusive and should be logged.

Use an extra button to save annotations (which may be done repeatedly)

this is extremely important.

Remove voice count down

It is now unnecessary for voice count down. Because now voice feedback always fired immediately after making a prediction

Add ethnicity to demography question

Dependabot can't parse your package.json

Dependabot couldn't parse the package.json found at /package.json.

You can mention @dependabot in the comments below to contact the Dependabot team.

Change default setting of sensor sampling rate to 100 Hz

Change voice feedback format

1. Shortened activity name
2. Only play activity name with short pause
3. Map confidence to voice volume

Add counter/timer for existing annotation/variation

Add a counter and a timer for the ongoing annotation/variation.

So that researcher can know when to stop the participant.

Save annotations first before stopping processors

Add simple interface for warm-up session

1. Add a new not mutually-exclusive warm-up annotation tag
2. Add a new mutually-exclusive "stretch" annotation tag
3. When warm-up tag is on,
   Override default model predictions, but instead playing prescriptive clips each time when the system wants to play something depending on the variation stage.
   It could be a random prediction during variation I and II, and could be always correct prediction during variation 0.

Add monitor to show sensor sampling rate

This is important to make sure data are coming in correctly.

Do not count the first window after "switching"

Log "variation" status

When on, "variation" should be logged to annotation.

Support AUDIO prediction feedback

Use Web speech synthesis API.

https://developer.mozilla.org/en-US/docs/Web/API/SpeechSynthesis

Disable voice feedback during "SYNC"

Support more demographic information

Age
Gender
Weight
Height
Rest heart rate
Rest blood pressure
Dress code
Shoe type

Use a hard configurable threshold to cut the predictions and use the same volume for all selected activities