Hands-Free Crowdsensing of Accessibility Barriers in Sidewalk Infrastructure: A Brain–Computer Interface Approach

Crowdsourced human sensing provides a critical complement to urban physical sensing networks to detect city infrastructure disrepairs and inform strategized maintenance. However, motivating public participation remains challenging due to the additional efforts required to engage with third-party platforms. This study builds upon the “citizen as sensor” notion but emphasizes an intuitive and effortless approach for people to contribute their observations and experiences regarding accessibility barriers in the physical world to cyber data systems. We develop a wearable electroencephalogram (EEG)-based brain–computer interface application to predict the presence or absence of environmental barriers and individuals’ reaction and categorization to various accessibility barrier conditions. To provide proof-of-concept, we first characterized the mental model for identifying environmental barriers based on Bayesian inference. A structured experiment was conducted, collecting empirical EEG data from 12 participants as they performed a sidewalk accessibility evaluation task. Based on correlations between extracted EEG power features and cognitive discriminability characterized by signal detection theory, we validated that EEG measures can reflect internal representations when individuals perceived a scenario as “inaccessible.” We then derived the decision threshold above which an accessibility barrier is reported from elicited EEG power patterns and based, on this threshold, estimated participants’ task performance using maximum a posteriori estimates. The accuracy of these estimates was confirmed by comparing them against actual performance data. Machine learning classifiers were then leveraged to enhance automatic detection and categorization of environmental features in terms of accessibility from individuals’ brain signals. Theoretically, theta activity in the temporal cortical region was found to dominate a memory retrieval-based cognitive process for identifying accessibility barriers. Practically, increased citizen participation in crowdsourcing is expected to provide valuable insights into mobility needs, inform infrastructure maintenance, and foster a sense of collaboration in creating inclusive urban environments.