Convolutional Algorithms for Real-Time Data Processing with Time Series on Smartphones for Navigation and Tracking
Most modern smartphones have various sensor types integrated and become a rich sensing platform for a wide range of applications including healthcare, sport, social media, environmental monitoring. In most of the applications, time-serial data such as the Inertial Measurement Unit (IMU) data is streamed continuously to the application and processed using a Convolutional Neural Network (CNN) model which typically uses the standard 2D convolution. However, there are many more convolution types (including depth-wise, dilated, transposed) that have not been exploited for real-time computation in an embedded device such as a smartphone.
To this end, this project aims at exploiting the various kind of CNN convolutions for IMU-based tracking.
The project will be divided into 2 main tasks:
- Study IMU-based tracking using Convolutional Neural Networks (CNN) with various type of convolutions
- Implement the models on a smartphone for a tracking application.
30% Theory, 50% Implementation, 20%Writing
Le Viet Duc, firstname.lastname@example.org, room ZI 5013