Advisor Meeting Time: 10:15 AM - 12:00 PM (Bi-Weekly)

General Meeting Time: Monday 4:20 PM - 8:00 PM, Wednesday 4:20PM - 8:00PM

Introduction

Assistive augmented vision technologies are becoming increasingly popular among the visually impaired. To function properly and to minimize user distraction, these technologies must determine a variety of context attributes of a user’s activities and his/her surrounding environment. Accurate step counting, as an important context attribute, is generally overlooked by augmented vision technologies. Due to factors such as independent movement of the head, an accelerometer located on the head presents a significantly noisier picture of the forces felt by the body when walking when compared to a more traditional location for a pedometer. As such, step counting using an accelerometer located on the head presents a unique challenge for step counting.
In this project, several data processing algorithms to extract a step count from accelerometer data will be examined. Several smoothing filters will be tested to help the step counting algorithm perform well on noisy, accelerometer data. Trade-offs will be explored between false positives and false negatives, as well as performance and accuracy.

Outcomes

Skills needed across the team:
Application development frameworks
Data visualization tools and techniques
Knowledge of smoothing and moving averages
Interest and ability to learn new digital technologies
Utilizing and creating RESTful APIs
Languages / tools:
Python, Java, MATLAB, Tableau

Skills students would develop:
Understanding of assertive technologies for the visually impaired
Multimodal data processing
Algorithm design
Context-awareness issues in mobile platforms
Unsupervised learning
Android programming
Design considerations and tradeoffs
Design thinking

Team Roles/Contact