**7. Conclusions**

We presented StressFoot, a pair of smart shoes that can sense stress unobtrusively by using a pressure sensitive insole and an IMU. Our prototype is capable of identifying acute stress and relaxation while sitting, such as performing office tasks. We identified four characteristics, which reflect foot pressure distributions, foot posture variations and foot tapping. Based on these features, we trained several machine learning models with 23 participants by using a leave-one*user*-out and validated this as a method to detect stress with an average accuracy of ∼85%. Then, with 11 additional participants, we demonstrated the replicability of our model with a similar overall accuracy of ∼87%. Finally, to evidence external validity, we conducted a field study with 10 participants, and evaluated the robustness of our models in an actual office setting. The outcome was that the computed stress level provided by our machine learning model correlates with the self-reported stress level with a coefficient of *r* = 0.79. We envision StressFoot to be an unobtrusive system capable of detecting the user's stress level on a daily basis. By drawing attention to the user's mental stress condition, such a system may already be able to contribute to an improvement in overall mental well-being in the future.

**Author Contributions:** D.S.E. conceptualised and Investigation as part of his PhD research. D.J.C.M. and S.N. contributed with overall supervision, review editing. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by *Assistive Augmentation* research gran<sup>t</sup> under the Entrepreneurial Universities (EU) initiative of New Zealand.

**Acknowledgments:** We acknowledge Ridmi Nimeshani Induruwa Bandarage for helping us in technical illustrations.

**Conflicts of Interest:** The authors declare no conflict of interest.
