After this next layer of reading, I feel I have moved from asking which sensors might be interesting to having a much clearer idea of what the sensing system actually needs to do.

The strongest pattern across the papers is that one physiological signal on its own is not enough. Heart rate, EDA, HRV and temperature can all change for many reasons, particularly when somebody is moving through a real environment. The more defensible approach is to combine physiological change with movement, context and the person's own baseline, and then use that information cautiously.

So my current thinking is moving towards:

personal baseline + physiological change + movement + airport context + previous responses → cautious interpretation → check whether this is a good moment to prompt → transparent check-in → traveller decides

This is a much clearer system than where I was a few weeks ago.

What I think the tracking needs to include

Based on the research so far, the strongest current sensing combination is:

EDA + PPG / pulse + accelerometry with skin temperature as a secondary contextual measure.

I am currently treating each signal as having a different role rather than expecting them all to tell me the same thing.

EDA

Useful for detecting changes in electrodermal activity and physiological arousal.

Recent comparison work suggests it can remain informative during movement, but it still does not tell me what the change means.

PPG / pulse

Useful for continuous pulse-rate trends.

The EmbracePlus validation study showed that wrist PPG can remain reasonably accurate during motion when accelerometry is used to help deal with artefacts.

Accelerometry

Possibly the most important contextual channel.

It can help identify whether a physiological change happened while the person was walking, rushing or relatively still, and can also help improve the quality of PPG interpretation.

I am now much less convinced that HRV / PRV should be a primary live trigger during movement. It can still be useful, but the driving-context research reinforces how cautious I need to be when interpreting HRV-derived measures outside controlled conditions.

Skin temperature remains interesting, particularly because both wrist and ring devices can measure it, but at the moment I see it as a secondary contextual signal rather than the main trigger.

The biggest change: personal baseline

The Oura study has made personal baseline feel essential rather than optional.

A universal threshold such as: If heart rate rises above X, offer support… does not fit the evidence.

The more defensible question is: Has something changed meaningfully compared with what is normal for this person? Even then, the system still does not know why the change happened. The baseline simply gives the sensing a more personalised reference point.

Ring or wrist?

At this stage, I think the hardware decision has narrowed to two genuinely plausible routes:

Oura-style ring

Pros

• very discreet and compact
• familiar consumer object rather than medical-looking equipment
• strong precedent for long-term personal baseline tracking
• pulse / inter-beat interval data
• HRV-related measures
• skin temperature
• movement / activity sensing
• potentially lower visual burden

Cons

• the Oura paper I reviewed did not include EDA
• finger comfort cannot be assumed for autistic users
• may be affected by hand use, luggage and grip
• gives me fewer physiological channels if EDA remains important
• the paper was useful for baseline methodology, not airport-specific performance

Empatica EmbracePlus-style wrist sensing

Pros

• EDA
• multiwavelength PPG / pulse
• 3-axis accelerometry
• skin temperature
• raw sensor access
• research-grade platform
• pulse-rate validation during movement
• wrist-based devices were reported as more comfortable than finger-electrode systems
• gives me all three of my current priority channels in one device

Cons

• physically larger and more visible than a ring
• correct fit matters for signal quality
• wrist pressure / texture could itself become a sensory issue
• hand and forearm movement can still introduce artefacts
• a research-grade wristband is not automatically the right final product form

My current choice

At this point, I think EmbracePlus is the stronger research / prototype platform for the next stage.

The main reason is not that it has “better” data in every situation. It is that it currently gives me the best match to the sensing architecture my research is pointing towards: EDA + PPG / pulse + accelerometry in one wearable, with raw data access and existing evidence that pulse measurement can remain useful during movement.

The ring is still extremely attractive as a form-factor direction, particularly because it is smaller and more discreet. If future work shows that EDA does not add enough value, I could imagine the project shifting back toward a ring-based architecture. But right now, removing EDA would feel premature.

Current working decision

For the next research and prototyping stage, I will treat EmbracePlus-style wrist sensing as the strongest current platform.

This is a decision about the research hardware and sensing architecture, not yet the final designed wearable form.

What the research is pointing towards

Across the papers, several ideas are becoming consistent.

1. More sensors do not automatically mean more understanding.

The Heng and Yavari comfort study showed that combining physiology, movement and environment is technically plausible, but real-world classification remained modest.

2. Movement has to be part of interpretation.

The driving-context study showed that real-world wearable classification became much weaker when evaluated properly across people, and accelerometry remained important.

3. Personal baseline matters.

The Oura work showed the value of participant-specific baseline correction and reinforced that absolute physiological thresholds are not defensible.

4. Wrist pulse sensing during movement is feasible, but imperfect.

The EmbracePlus validation study showed that PPG can remain useful during movement when accelerometry is used to help manage motion artefacts.

5. EDA still appears worth keeping.

The wearable comparison study suggested that EDA remained relatively useful during physical activity and that wrist-worn sensing offered a better comfort trade-off than finger-based electrodes.

6. Detection should lead to a question, not a conclusion.

The stress-logging study showed that wearable events can feasibly trigger user-confirmation prompts in everyday life.

7. Not every detected change should interrupt the person.

The JITAI paper showed the value of combining physiology with behaviour, environment and previous responses to decide whether a prompt is likely to be useful.

Midway conclusion

I am now much clearer about what biosensing is doing in the project. I am not trying to build a wearable that detects sensory overload.

I am designing a system that may notice a meaningful physiological change, understand whether the person is moving, consider what is happening in the airport, compare that moment with the person's own baseline and previous choices, and then decide whether it is appropriate to ask.

The traveller still decides what the change means. My current sensing direction is therefore:

EDA + PPG / pulse + accelerometry + HRV, interpreted relative to a personal baseline and combined with airport context.

My current hardware direction is:

EmbracePlus-style wrist sensing for research and prototyping, while keeping the final physical form open.

I now have enough evidence to move into the next part of the system: deciding exactly what contextual airport information the system needs, how prompts should work, what support options should follow, and how the interaction should behave when sensing is uncertain, wrong or unwanted.