An official website of the United States government
In today's world, AI systems need to make sense of large amounts of data as it unfolds in real-time, whether it's a video from surveillance and monitoring cameras, streams of egocentric footage, or sequences in other domains such as text or audio. The ability to break these continuous data streams into meaningful events, discover nested structures, and predict what might happen next at different levels of abstraction is crucial for applications ranging from passive surveillance systems to sensory-motor autonomous learning. However, most existing models rely heavily on large, annotated datasets with fixed data distributions and offline epoch-based training, which makes them impractical for handling the unpredictability and scale of dynamic real-world environments. This dissertation tackles these challenges by introducing a set of predictive models designed to process streaming data efficiently, segment events, and build sequential memory models without supervision or data storage. First, we present a single-layer predictive model that segments long, unstructured video streams by detecting temporal events and spatially localizing objects in each frame. The model is applied to wildlife monitoring footage, where it processes continuous, high-frame-rate video and successfully detects and tracks events without supervision. It operates in an online streaming manner to perform simultaneous training and inference without storing or revisiting the processed data. This approach alleviates the need for manual labeling, making it ideal for handling long-duration, real-world video footage. Building on this, we introduce STREAMER, a multi-layered architecture that extends the single-layer model into a hierarchical predictive framework. STREAMER segments events at different levels of abstraction, capturing the compositional structure of activities in egocentric videos. By dynamically adapting to various timescales, it creates a hierarchy of nested events and forms more complex and abstract representations of the input data. Finally, we propose the Predictive Attractor Model (PAM), which builds biologically plausible memory models of sequential data. Inspired by neuroscience, PAM uses sparse distributed representations and local learning rules to avoid catastrophic forgetting, allowing it to continually learn and make predictions without overwriting previous knowledge. Unlike many traditional models, PAM can generate multiple potential future outcomes conditioned on the same context, which allows for handling uncertainty in generative tasks. Together, these models form a unified framework of predictive learning that addresses multiple challenges in event understanding and temporal data analyses. By using prediction as the core mechanism, they segment continuous data streams into events, discover hierarchical structures across multiple levels of abstraction, learn semantic event representations, and model sequences without catastrophic forgetting.
more »
« less
This content will become publicly available on August 1, 2026
Enhancing Wearable Fall Detection System via Synthetic Data
Deep learning models rely heavily on extensive training data, but obtaining sufficient real-world data remains a major challenge in clinical fields. To address this, we explore methods for generating realistic synthetic multivariate fall data to supplement limited real-world samples collected from three fall-related datasets: SmartFallMM, UniMib, and K-Fall. We apply three conventional time-series augmentation techniques, a Diffusion-based generative AI method, and a novel approach that extracts fall segments from public video footage of older adults. A key innovation of our work is the exploration of two distinct approaches: video-based pose estimation to extract fall segments from public footage, and Diffusion models to generate synthetic fall signals. Both methods independently enable the creation of highly realistic and diverse synthetic data tailored to specific sensor placements. To our knowledge, these approaches and especially their application in fall detection represent rarely explored directions in this research area. To assess the quality of the synthetic data, we use quantitative metrics, including the Fréchet Inception Distance (FID), Discriminative Score, Predictive Score, Jensen–Shannon Divergence (JSD), and Kolmogorov–Smirnov (KS) test, and visually inspect temporal patterns for structural realism. We observe that Diffusion-based synthesis produces the most realistic and distributionally aligned fall data. To further evaluate the impact of synthetic data, we train a long short-term memory (LSTM) model offline and test it in real time using the SmartFall App. Incorporating Diffusion-based synthetic data improves the offline F1-score by 7–10% and boosts real-time fall detection performance by 24%, confirming its value in enhancing model robustness and applicability in real-world settings.
more »
« less
- Award ID(s):
- 2123749
- PAR ID:
- 10626690
- Publisher / Repository:
- MDPI
- Date Published:
- Journal Name:
- Sensors
- Volume:
- 25
- Issue:
- 15
- ISSN:
- 1424-8220
- Page Range / eLocation ID:
- 4639
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Falls are the second leading cause of unintentional injury deaths worldwide. While numerous wearable fall detection devices incorporating AI models have been developed, none of them are used successfully in a fall detection application running on commodity-based smartwatches in real time. The system misses some falls, and generates an annoying amount of False Positives for practical use. We have investigated and experimented with an LSTM model for fall detection on a smartwatch. Even though the LSTM model has high accuracy during offline testing, the good performance of offline LSTM models cannot be translated to the equivalence of real-time performance. Transformers, on the other hand, can learn long-sequence data and patterns intrinsic to the data due to their self-attention mechanism. This paper compares three variants of LSTM and two variants of Transformer models for learning fall patterns. We trained all models using fall and activity data from three datasets, and the real-time testing of the model was performed using the SmartFall App. Our findings showed that in the offline training, the CNN-LSTM model was better than the Transformer model for all the datasets. However, the Transformer is a preferable choice for deployment in real-time fall detection applications.more » « less
-
Oysters are a vital keystone species in coastal ecosystems, providing significant economic, environmental, and cultural benefits. As the importance of oysters grows, so does the relevance of autonomous systems for their detection and monitoring. However, current monitoring strategies often rely on destructive methods. While manual identification of oysters from video footage is non-destructive, it is time-consuming, requires expert input, and is further complicated by the challenges of the underwater environment. To address these challenges, we propose a novel pipeline using stable diffusion to augment a collected real dataset with photorealistic synthetic data. This method enhances the dataset used to train a YOLOv10-based vision model. The model is then deployed and tested on an edge platform; Aqua2, an Autonomous Underwater Vehicle (AUV), achieving a state-of-the-art 0.657 mAP@50 for oyster detection.more » « less
-
Camera-based systems are increasingly used for collecting information on intersections and arterials. Unlike loop controllers that can generally be only used for detection and movement of vehicles, cameras can provide rich information about the traffic behavior. Vision-based frameworks for multiple-object detection, object tracking, and near-miss detection have been developed to derive this information. However, much of this work currently addresses processing videos offline. In this article, we propose an integrated two-stream convolutional networks architecture that performs real-time detection, tracking, and near-accident detection of road users in traffic video data. The two-stream model consists of a spatial stream network for object detection and a temporal stream network to leverage motion features for multiple-object tracking. We detect near-accidents by incorporating appearance features and motion features from these two networks. Further, we demonstrate that our approaches can be executed in real-time and at a frame rate that is higher than the video frame rate on a variety of videos collected from fisheye and overhead cameras.more » « less
-
Falls in the elderly are associated with significant morbidity and mortality. While numerous fall detection devices incorporating AI and machine learning algorithms have been developed, no known smartwatch-based system has been used successfully in real-time to detect falls for elderly persons. We have developed and deployed a SmartFall system on a commodity-based smartwatch which has been trialled by nine elderly participants. The system, while being usable and welcomed by the participants in our trials, has two serious limitations. The first limitation is the inability to collect a large amount of personalized data for training. When the fall detection model, which is trained with insufficient data, is used in the real world, it generates a large amount of false positives. The second limitation is the model drift problem. This means an accurate model trained using data collected with a specific device performs sub-par when used in another device. Therefore, building one model for each type of device/watch is not a scalable approach for developing smartwatch-based fall detection system. To tackle those issues, we first collected three datasets including accelerometer data for fall detection problem from different devices: the Microsoft watch (MSBAND), the Huawei watch, and the meta-sensor device. After that, a transfer learning strategy was applied to first explore the use of transfer learning to overcome the small dataset training problem for fall detection. We also demonstrated the use of transfer learning to generalize the model across the heterogeneous devices. Our preliminary experiments demonstrate the effectiveness of transfer learning for improving fall detection, achieving an F1 score higher by over 10% on average, an AUC higher by over 0.15 on average, and a smaller false positive prediction rate than the non-transfer learning approach across various datasets collected using different devices with different hardware specifications.more » « less
