No abstract

No abstract

This examines the use of very low frequency (VLF) electromagnetic signals with artificial neural networks (ANN) to estimate two‐dimensional location. The VLF source positions and type of signals are not known in advance. VLF signals were collected from the environment using a mobile antenna while the user manually tagged the true location. Data were collected over several months. The signals were divided into time segments, and features were generated for the ANN from the spectral power of the signals over time. The ANN uses long short‐term memory (LSTM) layers and fully connected layers and was trained using the time‐series spectral features of the VLF signal and the true location data. The position estimates are an indoor locale‐classification problem. Our task was to determine which of 50 rooms or hallway segments the sensor was in for a given building. The system achieved an average accuracy over 76% in determining discrete rooms or hallways. The approach achieved these exceptional accuracies with very little a priori knowledge of the VLF signals using only supervised learning.

is the Director of the Autonomy and Navigation Technology (ANT) Center at the Air Force Institute of Technology, where he is also an Associate Professor of Electrical Engineering. He has been involved in navigation-related research for over 25 years.

No abstract