arXiv subjects

Thomas Vaessen

Publications and source records attributed to Thomas Vaessen.

4 recordsLinked to original sources

arXiv2025

Machine Learning-based Context-Aware EMAs: An Offline Feasibility Study

Mobile health (mHealth) systems help researchers monitor and care for patients in real-world settings. Studies utilizing mHealth applications use Ecological Momentary Assessment (EMAs), passive sensing, and contextual features to develop emotion recognition models, which rely on EMA responses as ground truth. Due to this, it is crucial to consider EMA compliance when conducting a successful mHealth study. Utilizing machine learning is one approach that can solve this problem by sending EMAs based on the predicted likelihood of a response. However, literature suggests that this approach may lead to prompting participants more frequently during emotions associated with responsiveness, thereby narrowing the range of emotions collected. We propose a multi-objective function that utilizes machine learning to identify optimal times for sending EMAs. The function identifies optimal moments by combining predicted response likelihood with model uncertainty in emotion predictions. Uncertainty would lead the function to prioritize time points when the model is less confident, which often corresponds to underrepresented emotions. We demonstrate that this objective function would result in EMAs being sent when participants are responsive and experiencing less commonly observed emotions. The evaluation is conducted offline using two datasets: (1) 91 spousal caregivers of individuals with Alzheimer's Disease and Related dementias (ADRD), (2) 45 healthy participants. Results show that the multi-objective function tends to be higher when participants respond to EMAs and report less commonly observed emotions. This suggests that using the proposed objective function to guide EMA delivery could improve receptivity rates and capture a broader range of emotions.

cs.HC↗

arXiv2023

Towards the Understanding of Receptivity and Affect in EMAs using Physiological based Machine Learning Method: Analysis of Receptivity and Affect

As mobile health (mHealth) studies become increasingly productive due to the advancements in wearable and mobile sensor technology, our ability to monitor and model human behavior will be constrained by participant receptivity. The reliance on subjective responses for health constructs poses challenges, especially in populations with lower receptivity rates. Researchers have proposed machine-learning approaches to optimize survey timing and delivery to address this. However, there are concerns regarding potential biases or unintended influences on participant responses. Our study delves into factors impacting receptivity to ecological momentary assessments (EMA) in a 10-day mHealth study, exploring physiological relationships indicative of receptivity and affect. Utilizing data from 45 participants with wearable devices measuring various biometrics, we employ unsupervised (k-means clustering) and supervised (Random Forest and Neural Networks) machine learning methods to infer affect during non-responses. Findings reveal that triggering EMAs based on a receptivity model reduces reported negative affect by over 3 points (0.29 standard deviations). The predicted affect during non-responses exhibits a bimodal distribution, suggesting more frequent initiation during states of higher positive emotions. The study underscores a clear relationship between affect and receptivity, impacting mHealth study efficacy, especially those using machine learning for EMA triggering. Therefore, we propose a smart trigger that promotes EMA receptivity without influencing affect during sampled time points as future work.

cs.HC↗

arXiv2022

More to Less (M2L): Enhanced Health Recognition in the Wild with Reduced Modality of Wearable Sensors

Accurately recognizing health-related conditions from wearable data is crucial for improved healthcare outcomes. To improve the recognition accuracy, various approaches have focused on how to effectively fuse information from multiple sensors. Fusing multiple sensors is a common scenario in many applications, but may not always be feasible in real-world scenarios. For example, although combining bio-signals from multiple sensors (i.e., a chest pad sensor and a wrist wearable sensor) has been proved effective for improved performance, wearing multiple devices might be impractical in the free-living context. To solve the challenges, we propose an effective more to less (M2L) learning framework to improve testing performance with reduced sensors through leveraging the complementary information of multiple modalities during training. More specifically, different sensors may carry different but complementary information, and our model is designed to enforce collaborations among different modalities, where positive knowledge transfer is encouraged and negative knowledge transfer is suppressed, so that better representation is learned for individual modalities. Our experimental results show that our framework achieves comparable performance when compared with the full modalities. Our code and results will be available at https://github.com/compwell-org/More2Less.git.

cs.LG↗

arXiv2021

Modality Fusion Network and Personalized Attention in Momentary Stress Detection in the Wild

Multimodal wearable physiological data in daily life have been used to estimate self-reported stress labels. However, missing data modalities in data collection makes it challenging to leverage all the collected samples. Besides, heterogeneous sensor data and labels among individuals add challenges in building robust stress detection models. In this paper, we proposed a modality fusion network (MFN) to train models and infer self-reported binary stress labels under both complete and incomplete modality conditions. In addition, we applied personalized attention (PA) strategy to leverage personalized representation along with the generalized one-size-fits-all model. We evaluated our methods on a multimodal wearable sensor dataset (N=41) including galvanic skin response (GSR) and electrocardiogram (ECG). Compared to the baseline method using the samples with complete modalities, the performance of the MFN improved by 1.6% in f1-scores. On the other hand, the proposed PA strategy showed a 2.3% higher stress detection f1-score and approximately up to 70% reduction in personalized model parameter size (9.1 MB) compared to the previous state-of-the-art transfer learning strategy (29.3 MB).

eess.SP↗