Tekscan specializes in paper-thin force and pressure sensors as well as data acquisition and analysis software. Their medical division primarily focuses on gait analysis products for researchers and clinicians.
Tekscan's custom software development offerings to customers have previously been difficult to use and maintain. Along with several other engineers, I helped to co-develop a revamped software development kit (SDK) that enables users to write custom applications using our hardware and data files. This SDK is based on the .NET framework and written in C#, making it compatible with MATLAB, LabVIEW, C#, and other popular computing platforms.
I have been primarily responsible for ensuring compatibility with each of our these supported platforms, writing demo applications and sample code, documentation, and testing on a variety of platforms. I have also developed an application to check for product updates on Tekscan's live updates server. I am very excited about this product because it promises to provide additional power to customers and expedite internal custom software development.
Unfortunately, I cannot provide details publicly at this time.
As part of a larger custom software project for a third-party company, I have led an initial investigation of the use of neural networks and other machine learning algorithms to estimate subject shoe size from pressure sensor data alone. This is a non-trivial problem because of inconsistencies in the sensor measurements and variations with user weight and sensor sensitivity.
We have collected sample data from many subjects with precisely measured shoe sizes and hope to estimate this shoe size within a half size by implementing appropriate machine learning algorithms. Using MATLAB, I have implemented a neural network fitting algorithm with a variable number of layers and hidden units that has improved upon the previous approach. As this project moves forward, we hope to further improve performance and investigate potential enhancements.