Edison Thomaz

Dr. Edison Thomaz

orcid

Research Focus & Metrics

Academic Career Length

11Years (2016-Present)

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

Representative Publications

View All Publications
8 November 2021 in JMIR Serious Games
Verified
Journal Article
Open access
KRKavita Radhakrishnan
Kavita Radhakrishnan
CJChristine Julien
Christine Julien
TBTom Baranowski
Tom Baranowski
MOMatthew O'HAir
Matthew O'HAir
GLGrace Lee
Grace Lee
MKMiyong Kim
Miyong Kim
Background: Poor self-management of heart failure (HF) contributes to devastating health consequences. Our innovative sensor-controlled digital game (SCDG) integrates data from sensors to trigger game rewards, progress, and feedback based on the real-time behaviors of individuals with HF. Objective: The aim of this study is to compare daily weight monitoring and physical activity behavior adherence by older adults using an SCDG intervention versus a sensors-only intervention in a feasibility randomized controlled trial. Methods: English-speaking adults with HF aged 55 years or older who owned a smartphone and could walk unassisted were recruited from Texas and Oklahoma from November 2019 to August 2020. Both groups were given activity trackers and smart weighing scales to track behaviors for 12 weeks. The feasibility outcomes of recruitment, retention, intervention engagement, and satisfaction were assessed. In addition to daily weight monitoring and physical activity adherence, the participants’ knowledge, functional status, quality of life, self-reported HF behaviors, motivation to engage in behaviors, and HF-related hospitalization were also compared between the groups at baseline and at 6, 12, and 24 weeks. Results: A total of 38 participants with HF—intervention group (IG; 19/38, 50%) and control group (CG; 19/38, 50%)—were enrolled in the study. Of the 38 participants, 18 (47%) were women, 18 (47%) were aged 65 years or older, 21 (55%) had been hospitalized with HF in the past 6 months, and 29 (76%) were White. Furthermore, of these 38 participants, 31 (82%)—IG (15/19, 79%) and CG (16/19, 84%)—had both weight monitoring and physical activity data at the end of 12 weeks, and 27 (71%)—IG (14/19, 74%) and CG (13/19, 68%)—participated in follow-up assessments at 24 weeks. For the IG participants who installed the SCDG app (15/19, 79%), the number of days each player opened the game app was strongly associated with the number of days the player engaged in weight monitoring (r=0.72; P=.04) and the number of days with physical activity step data (r=0.9; P<.001). The IG participants who completed the satisfaction survey (13/19, 68%) reported that the SCDG was easy to use. Trends of improvement in daily weight monitoring and physical activity in the IG, as well as within-group improvements in HF functional status, quality of life, knowledge, self-efficacy, and HF hospitalization in both groups, were observed in this feasibility trial. Conclusions: Playing an SCDG on smartphones was feasible and acceptable for older adults with HF for motivating daily weight monitoring and physical activity. A larger efficacy trial of the SCDG intervention will be needed to validate trends of improvement in daily weight monitoring and physical activity behaviors. Trial Registration: ClinicalTrials.gov NCT03947983; https://clinicaltrials.gov/ct2/show/NCT03947983

12 September 2022 in Sensors
Verified
Journal Article
Open access
PDF file
RARebecca Adaimi
Rebecca Adaimi
ETEdison Thomaz
Edison Thomaz
Continual learning (CL), also known as lifelong learning, is an emerging research topic that has been attracting increasing interest in the field of machine learning. With human activity recognition (HAR) playing a key role in enabling numerous real-world applications, an essential step towards the long-term deployment of such systems is to extend the activity model to dynamically adapt to changes in people’s everyday behavior. Current research in CL applied to the HAR domain is still under-explored with researchers exploring existing methods developed for computer vision in HAR. Moreover, analysis has so far focused on task-incremental or class-incremental learning paradigms where task boundaries are known. This impedes the applicability of such methods for real-world systems. To push this field forward, we build on recent advances in the area of continual learning and design a lifelong adaptive learning framework using Prototypical Networks, LAPNet-HAR, that processes sensor-based data streams in a task-free data-incremental fashion and mitigates catastrophic forgetting using experience replay and continual prototype adaptation. Online learning is further facilitated using contrastive loss to enforce inter-class separation. LAPNet-HAR is evaluated on five publicly available activity datasets in terms of its ability to acquire new information while preserving previous knowledge. Our extensive empirical results demonstrate the effectiveness of LAPNet-HAR in task-free CL and uncover useful insights for future challenges.

17 March 2019
Verified
Conference Paper
Open access
KCKeum San Chun
Keum San Chun
ASAshley B. Sanders
Ashley B. Sanders
RARebecca Adaimi
Rebecca Adaimi
NSNecole Streeper
Necole Streeper
DCDavid E. Conroy
David E. Conroy
ETEdison Thomaz
Edison Thomaz
Over the last decade, advances in mobile technologies have enabled the development of intelligent systems that attempt to recognize and model a variety of health-related human behaviors. While automated dietary monitoring based on passive sensors has been an area of increasing research activity for many years, much less attention has been given to tracking fluid intake. In this work, we apply an adaptive segmentation technique on a continuous stream of inertial data captured with a practical, off-the-shelf wrist-mounted device to detect fluid intake gestures passively. We evaluated our approach in a study with 30 participants where 561 drinking instances were recorded. Using a leave-one-participant-out (LOPO), we were able to detect drinking episodes with 90.3% precision and 91.0% recall, demonstrating the generalizability of our approach. In addition to our proposed method, we also contribute an anonymized and labeled dataset of drinking and non-drinking gestures to encourage further work in the field.

23 June 2017
Verified
Conference Paper
Open access
ETEdison Thomaz
Edison Thomaz
ABAbdelkareem Bedri
Abdelkareem Bedri
TPTemiloluwa Prioleau
Temiloluwa Prioleau
IEIrfan Essa
Irfan Essa
GAGregory D. Abowd
Gregory D. Abowd
Motivated by health applications, eating detection with off-the-shelf devices has been an active area of research. A common approach has been to recognize and model individual intake gestures with wrist-mounted inertial sensors. Despite promising results, this approach is limiting as it requires the sensing device to be worn on the hand performing the intake gesture, which cannot be guaranteed in practice. Through a study with 14 participants comparing eating detection performance when gestural data is recorded with a wrist-mounted device on (1) both hands, (2) only the dominant hand, and (3) only the non-dominant hand, we provide evidence that a larger set of arm and hand movement patterns beyond food intake gestures are predictive of eating activities when L1 or L2 normalization is applied to the data. Our results are supported by the theory of asymmetric bimanual action and contribute to the field of automated dietary monitoring. In particular, it shines light on a new direction for eating activity recognition with consumer wearables in realistic settings.

22 July 2022 in Building and Environment
Verified
Journal Article
Closed access
HFHagen Fritz
Hagen Fritz
SBSepehr Bastami
Sepehr Bastami
CLCalvin Lin
Calvin Lin
KNKingsley Nweye
Kingsley Nweye
TTTung To
Tung To
ZNZoltan Nagy
Zoltan Nagy

27 September 2023 in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
Verified
Journal Article
Open access
DLDawei Liang
Dawei Liang
AZAlice Zhang
Alice Zhang
ETEdison Thomaz
Edison Thomaz
Understanding social interactions is relevant across many domains and applications, including psychology, behavioral sciences, human computer interaction, and healthcare. In this paper, we present a practical approach for automatically detecting face-to-face conversations by leveraging the acoustic sensing capabilities of an off-the-shelf, unmodified smartwatch. Our proposed framework incorporates feature representations extracted from different neural network setups and shows the benefits of feature fusion. The framework does not require an acoustic model specifically trained to the speech of the individual wearing the watch or of those nearby. We evaluate our framework with 39 participants in 18 homes in a semi-naturalistic study and with four participants in free living, obtaining an F1 score of 83.2% and 83.3% respectively for detecting user's conversations with the watch. Additionally, we study the real-time capability of our framework by deploying a system on an actual smartwatch and discuss several strategies to improve its practicality in real life. To support further work in this area by the research community, we also release our annotated dataset of conversations.