Gnss-receiver interference detection using deep learning

Abstract

Systems and methods are described for classification of interference for GNSS receivers. One or more neural networks are utilized to classify RF signal data received by a GNSS receiver. The classification associates the RF signal data with an RF environment. Appropriate interference mitigation techniques can be implemented by the receiver based on the classification.

What is claimed is: 
     
         1 . A GNSS processor architecture for processing GNSS receiver signal data, the processing architecture comprising:
 a processor; and   a memory unit in communication with the processor via a communication infrastructure and configured to store processor-readable instructions;
 wherein, when executed by the processor, the processor-readable instructions cause the processor to: 
 receive RF signal data associated with a class of RF environment; 
 provide the RF signal data to a neural network for classification of the RF signal data as belonging to a pre-defined type of RF environment; 
 using the neural network, classify the RF signal data as belonging to a pre-defined type of RF environment; and 
 apply an interference mitigation technique corresponding to the type of RF environment that has been classified. 
   
     
     
         2 . The GNSS processor architecture of  claim 1 , wherein the neural network comprises an artificial neural network (ANN). 
     
     
         3 . The GNSS processor architecture of  claim 1 , wherein the neural network comprises a convolutional neural network (CNN). 
     
     
         4 . The GNSS processor architecture of  claim 1 , wherein the RF signal data is classified as belonging to one of three RF environments, no interference, in-band interference, and out-of-band interference. 
     
     
         5 . The GNSS processor architecture of  claim 1 , wherein the neural network comprises a first layer used to input power spectral density (PSD) images. 
     
     
         6 . The GNSS processor architecture of  claim 5 , wherein the PSD images are configured in a 128×128 pixels format. 
     
     
         7 . The GNSS processor architecture of  claim 5 , wherein the neural network comprises eight convolutional layers configured to process the PSD images. 
     
     
         8 . The GNSS processor architecture of  claim 7 , wherein a 6×6 kernel is used for feature extraction using max pooling. 
     
     
         9 . The GNSS processor architecture of  claim 7 , wherein a last convolution layer is connected to a fully-connected layer with 32 neurons. 
     
     
         10 . The GNSS processor architecture of  claim 9 , wherein the neural network comprises an output layer of three (3) neurons configured to output classification determinations of one of three interference environments, using a soft max activation function. 
     
     
         11 . The GNSS processor architecture of  claim 1 , wherein the neural network comprises a recurrent neural network (RNN). 
     
     
         12 . The GNSS processor architecture of  claim 11 , wherein the RNN comprises a Long Short-Term Memory (LSTM) RNN. 
     
     
         13 . The GNSS processor architecture of  claim 1 , wherein the neural network comprises an autoencoder. 
     
     
         14 . The GNSS processor architecture of  claim 1 , wherein the neural network comprises a Restricted Boltzmann Machine (RBM).