SJSU ScholarWorks, San José State University's institutional repository, makes the research and creative works of faculty, staff and students available for the world to discover and use.
Open access scholarship can be read and cited by a global audience of scholars, students, and industry professionals. Each year, users from more than 200 countries download over 1.4 million items from SJSU ScholarWorks.
The SJSU Library preserves materials in SJSU ScholarWorks as part of our digital collections. SJSU ScholarWorks contains a broad range of scholarship, creative work, archival collections, and more:
Applications of ubiquitous computing, including health monitoring, sports analytics, and ambient-assisted living, rely on Human Activity Recognition (HAR) using wearable sensors. However, model robustness is challenged by missing sensor values, class imbalance, inter-subject variability, and temporal noise. This work proposes a complete HAR pipeline that addresses these challenges through sampling, time-series augmentation, dynamic feature handling, and GAN-PCA-based imputation. Built on the DeepSense architecture, the model integrates convolutional feature extraction with bi-GRUs for temporal modeling. The system is evaluated using 5-fold cross-validation, subject-aware holdout, and LOSEO strategies on the Opportunity dataset. Results demonstrate consistent accuracy across folds and strong generalization to unseen sessions and subjects, significantly outperforming baseline models. This study highlights the effectiveness of hybrid deep learning approaches in real-world HAR scenarios
VisionMate is a web application that generates captions for camera-captured images. It is designed
to assist users with visual impairments by converting visual input into spoken and written text. The
application uses the GIT-base model from Hugging Face, which processes the image and returns a
descriptive caption. Users can take a picture using the device camera—either via webcam on
desktop or the native camera interface on mobile. The app provides audio output using the
SpeechSynthesis API and uses full-screen tap interaction to simplify accessibility.
The frontend is implemented in React.js, and the backend is built with FastAPI. The backend calls
Hugging Face’s Inference API to perform model inference without loading large models locally,
reducing memory usage during deployment. On average, captions are generated in 5 to 8 seconds.
The GIT-base model was selected after comparative testing with BLIP-base, BLIP-large, and GIT-
large. Testing was conducted on Chrome, Safari, and Firefox browsers, using devices such as the
MacBook Pro (M1) and iPhone 15.
This report outlines the system architecture, model comparisons, deployment on Vercel (frontend)
and Render (backend), and evaluation of performance across speed, model accuracy, and device and
browser compatibility.
In-depth understanding of the complexity of daily human activities is crucial for building responsive health monitoring and assistive technologies. However, limited research has focused on distinguishing activities based on their involvement level, as most existing work classifies only the type of activity performed. In this thesis, we address this gap by proposing a method to classify human activities as either simple or complex using sensor data from the Opportunity dataset. We define complex activities as those involving object interactions or multiple coordinated movements (e.g., drinking from a cup, cleaning a table), and simple activities as static or low-effort postures (e.g., standing, sitting). Based on domain-specific heuristics, we label each time window and extract both time-domain and frequency-domain features. We evaluate multiple models, including a hybrid deep learning architecture combining Convolutional Neural Networks (CNNs) and Transformers, trained on multi-sensor data across varied window lengths. To ensure robustness and generalization, we apply Leave-One-User-Out (LOUO) and Leave-One-Episode- Out (LOEO) evaluation. Our results show that the proposed approach reliably distinguishes between simple and complex activities, outperforming traditional classifiers and offering new directions for fine-grained activity recognition in real-world environments.
In modern car manufacturing, collaborative robots (cobots) work with human operators during shared workcell interactions to maximize production speed and flexibility. Collaboration between humans and robots is safe and effective only when operator intent recognition via a single wrist-worn inertial measurement unit (IMU) is accurate and low-latency. This thesis develops an IMU-only intent recognition pipeline, and is evaluated on three datasets: the public OPPORTUNITY dataset, the Sony Smartwatch Gesture dataset and a custom Samsung Galaxy watch 6 dataset. The proposed framework leverages five step sequence-to-label problems which are stepwise posed as data streams transforming raw IMU data into trainable tensors. For that purpose, a deterministic seven-step transformation consists of transport-agnostic ingestion, gap interpolation, low-pass filtering, and on-the-fly label-uncorrupted znormalisation. Z-normalised epochs of 100 frames at sampling frequency 50 Hz are fed to five sequence-deep architectures. The BiLSTM, GRU, 1D CNN, CNN+LSTM, and a four-block Temporal Convolutional Network TCN were trained and evaluated under stratified 80/20, Leave-One-Subject-Out (LOSO), Leave-One-Experiment-Out (LOEO), and stratified K-fold validation—all separately on each dataset. The TCN consistently achieves the best trade-off between accuracy and efficiency: a macro-F1 score of 0.923 on the OPPORTUNITY 80/20 split, 0.967 under LOSO on the Sony corpus, and 0.427 under LOSO on OPPORTUNITY using only 200k parameters while sustaining sub-20 ms inference on standard CPUs. A Bayesian hyperparameter search on the SJSU HPC cluster demonstrates a 40% reduction in tuning time, showcasing scalable reproducibility across 1,232 CPU cores and multiple GPU nodes.
The Warm Springs BART Extension Transit Village is a transit oriented development proposed for a site in south Fremont, California. Residents of Alameda County approved a measure to extend the Bay Area Rapid Transit (BART) line from Fremont to Warm Springs. To date, a Supplemental Environmental Impact Report (SEIR) and a Specific Plan have been prepared for the station area.
According to the SEIR, BART has proposed a parking lot designed around the station area to accommodate around 2040 on-site parking spaces.
The main objective of this study is to provide a design alternative for the City of Fremont, through the concept of a transit village using New Urbanism principles that is far more functional and economical than an open parking lot.
The site at present is zoned for industrial uses and has very little development surrounding the station. This design presents a unique opportunity to change this nearly “dead zone” to a livelier and vibrant community through the implementation of a transit oriented development (TOD) which is development that primarily occurs within walking distance from a transit stop or hub (1/2 mile radius) using the New Urbanism principles that encourage high density, mixed use and a pedestrian friendly environment.
The study area for which the TOD is proposed in this report is approximately 14 mile radius surrounding the BART extension in Warm Springs. While the BART Station acts as the focal point in the neighborhood, it aims at connecting the rest of the community through an integrated street network that is both walkable and pedestrian friendly without eliminating the automobile. The block pattern used for the design maximizes the potential for a variety of mixed uses including housing, retail and offices, and acts as a gateway to the hillside suburban development of Fremont. It is designed to present a unique identity to the neighborhood and serves as a medium to maximize future development potential for the region.
The study that follows illustrates the design of a Transit Village using New Urbanism principles, explores the need for TOD as a planning tool, reviews TOD policies in the state of California and discusses the importance of the design more specifically for the City of Fremont and for the Warm Springs BART extension. The study concludes by suggesting design recommendations and land use changes to the site for the successful implementation of the design proposal.
