Bryan Minor

Dr. Bryan Minor

Center for Advanced Studies in Adaptive Systems, Voiland College of Engineering and Architecture, Washington State University, PO Box 642250, Pullman, WA 99164-2250, USA

Research Focus & Metrics

Academic Career Length

10Years (2017-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
11 September 2017 in IEEE Transactions on Knowledge and Data Engineering
Verified
Journal Article
Open access
Bryan MinorBryan Minor
Bryan Minor
JDJanardhan Rao Doppa
Janardhan Rao Doppa
Diane CookDiane Cook
Diane Cook
Recent progress in Internet of Things (IoT) platforms has allowed us to collect large amounts of sensing data. However, there are significant challenges in converting this large-scale sensing data into decisions for real-world applications. Motivated by applications like health monitoring and intervention and home automation we consider a novel problem called Activity Prediction, where the goal is to predict future activity occurrence times from sensor data. In this paper, we make three main contributions. First, we formulate and solve the activity prediction problem in the framework of imitation learning and reduce it to a simple regression learning problem. This approach allows us to leverage powerful regression learners that can reason about the relational structure of the problem with negligible computational overhead. Second, we present several metrics to evaluate activity predictors in the context of real-world applications. Third, we evaluate our approach using real sensor data collected from 24 smart home testbeds. We also embed the learned predictor into a mobile-device-based activity prompter and evaluate the app for nine participants living in smart homes. Our results indicate that our activity predictor performs better than the baseline methods, and offers a simple approach for predicting activities from sensor data.

4 July 2025 in IEEE Journal of Biomedical and Health Informatics
Verified
Journal Article
Closed access
Bryan MinorBryan Minor
Bryan Minor
CGColin Greeley
Colin Greeley
RHRyan Holder
Ryan Holder
BTBrian Thomas
Brian Thomas
Lawrence HolderLawrence Holder
Lawrence Holder
Diane CookDiane Cook
Diane Cook
Human activity recognition (HAR) from wearable sensor data traditionally identifies atomic movements (e.g., sit, stand, walk). However, many medical fields require recognizing functional activities—higher-level, goal-directed behaviors (e.g., errands, socialize, work). Functional activity recognition is critical for cognitive health assessment, rehabilitation, post-surgical recovery, and chronic disease management, yet remains largely unexplored due to its inherent complexity and variability for in-the-wild settings. This work addresses these challenges by investigating methods for functional HAR and introducing a novel approach that augments feature representations with feature token-transformer embeddings to improve classification performance. We compare a range of machine learning and deep learning methods, analyzing their ability to generalize across a diverse population. Additionally, we present ArWISE, a large-scale functional activity dataset collected longitudinally from n = 503 participants, consisting of over 32 million labeled points. Our experiments demonstrate the advantages of incorporating feature embeddings into functional HAR models, particularly in handling real-world variability and data sparsity. By bridging the gap between atomic movement recognition and functional behavior modeling, this work lays the foundation for more advanced, behavior-aware applications in digital health and humancentered AI.

10 April 2025 in JMIR mHealth and uHealth
Verified
Journal Article
Open access
Diane CookDiane Cook
Diane Cook
AWAiden Walker
Aiden Walker
Bryan MinorBryan Minor
Bryan Minor
CLCatherine Luna
Catherine Luna
SFSarah Tomaszewski Farias
Sarah Tomaszewski Farias
MSMaureen Schmitter-Edgecombe
Maureen Schmitter-Edgecombe
Background: Ecological momentary assessment (EMA) offers an effective method to collect frequent, real-time data on an individual’s well-being. However, challenges exist in response consistency, completeness, and accuracy. Objective: This study examines EMA response patterns and their relationship with sensed behavior for data collected from diverse studies. We hypothesize that EMA response rate (RR) will vary with prompt time of day, number of questions, and behavior context. In addition, we postulate that response quality will decrease over the study duration and that relationships will exist between EMA responses, participant demographics, behavior context, and study purpose. Methods: Data from 454 participants in 9 clinical studies were analyzed, comprising 146,753 EMA mobile prompts over study durations ranging from 2 weeks to 16 months. Concurrently, sensor data were collected using smartwatch or smart home sensors. Digital markers, such as activity level, time spent at home, and proximity to activity transitions (change points), were extracted to provide context for the EMA responses. All studies used the same data collection software and EMA interface but varied in participant groups, study length, and the number of EMA questions and tasks. We analyzed RR, completeness, quality, alignment with sensor-observed behavior, impact of study design, and ability to model the series of responses. Results: The average RR was 79.95%. Of those prompts that received a response, the proportion of fully completed response and task sessions was 88.37%. Participants were most responsive in the evening (82.31%) and on weekdays (80.43%), although results varied by study demographics. While overall RRs were similar for weekday and weekend prompts, older adults were more responsive during the week (an increase of 0.27), whereas younger adults responded less during the week (a decrease of 3.25). RR was negatively correlated with the number of EMA questions (r=−0.433, P<.001). Additional correlations were observed between RR and sensor-detected activity level (r=0.045, P<.001), time spent at home (r=0.174, P<.001), and proximity to change points (r=0.124, P<.001). Response quality showed a decline over time, with careless responses increasing by 0.022 (P<.001) and response variance decreasing by 0.363 (P<.001). The within-study dynamic time warping distance between response sequences averaged 14.141 (SD 11.957), compared with the 33.246 (SD 4.971) between-study average distance. ARIMA (Autoregressive Integrated Moving Average) models fit the aggregated time series with high log-likelihood values, indicating strong model fit with low complexity. Conclusions: EMA response patterns are significantly influenced by participant demographics and study parameters. Tailoring EMA prompt strategies to specific participant characteristics can improve RRs and quality. Findings from this analysis suggest that timing EMA prompts close to detected activity transitions and minimizing the duration of EMA interactions may improve RR. Similarly, strategies such as gamification may be introduced to maintain participant engagement and retain response variance.

6 January 2025 in Educational Gerontology
Verified
Journal Article
Closed access
LWLisa Ann Kirk Wiese
Lisa Ann Kirk Wiese
CLCatherine Luna
Catherine Luna
Diane CookDiane Cook
Diane Cook
Bryan MinorBryan Minor
Bryan Minor
MSMaureen Schmitter-Edgecombe
Maureen Schmitter-Edgecombe
Rural, older racially and ethnically diverse residents in rural areas in the southern U.S. have reported a strong sense of community, and desire to pursue healthy behaviors that promote aging in place. However, the ability to monitor health habits through wireless fitness devices is hindered in this underresourced area by lack of access and reduced literacy of mobile technology. A partnership between high school students, local faith-based health educators and older adults was developed, with the hypothesis that this approach would result in acceptance and use of the technology to monitor and improve health. Ecological momentary assessment (EMA) was used to administer health survey questions (e.g., In the past two hours, how much social contact have you had?) through the smartwatch 4x per day in this pilot study (n = 33 older adults). Paper and pencil measures included brief screening for cognition, health literacy, and technology competence. Feasibility rates and Pearson correlational analyses were used to examine relationships between screening measures and smartwatch use. Results included that 91% of older adults responded to the smartwatch prompts (response rate = 77.80%). Response rates were unrelated to health literacy, technology self-efficacy, self-reported cognition, education, age, or rural living. Social contact was positively related to physical activity, being mentally engaged, and “mind as sharp as usual.” The training process and smartwatches functioned for participants across a continuum of skill levels with technology and health literacy. Strengths, barriers, and opportunities for improvement to this intergenerational approach, as reported by participants and researchers, are shared.

1 September 2019
Verified
Conference Paper
Closed access
SMSeyed Iman Mirzadeh
Seyed Iman Mirzadeh
JAJessica Ardo
Jessica Ardo
RFRamin Fallahzadeh
Ramin Fallahzadeh
Bryan MinorBryan Minor
Bryan Minor
LELorraine S. Evangelista
Lorraine S. Evangelista
HGHassan Ghasemzadeh
Hassan Ghasemzadeh
While inferring human activities from sensors embedded in mobile devices using machine learning algorithms has been studied, current research relies primarily on sensor data that are collected in controlled settings often with healthy individuals. Currently, there exists a gap in research about how to design activity recognition models based on sensor data collected with chronically-ill individuals and in free-living environments. In this paper, we focus on a situation where free-living activity data are collected continuously, activity vocabulary (i.e., class labels) are not known as a priori, and sensor data are annotated by end-users through an active learning process. By analyzing sensor data collected in a clinical study involving patients with cardiovascular disease, we demonstrate significant challenges that arise while inferring physical activities in uncontrolled environments. In particular, we observe that activity labels that are distinct in syntax can refer to semantically-identical behaviors, resulting in a sparse label space. To construct a meaningful label space, we propose LabelMerger, a framework for restructuring the label space created through active learning in uncontrolled environments in preparation for training activity recognition models. LabelMerger combines the semantic meaning of activity labels with physical attributes of the activities (i.e., domain knowledge) to generate a flexible and meaningful representation of the labels. Specifically, our approach merges labels using both word embedding techniques from the natural language processing domain and activity intensity from the physical activity research. We show that the new representation of the sensor data obtained by LabelMerger results in more accurate activity recognition models compared to the case where original label space is used to learn recognition models.