NLOS Identification in WLANs Using Deep LSTM with CNN Features.

Area of Science:

• Wireless Communications
• Machine Learning
• Signal Processing

Background:

• Optimizing location-based services in wireless communications relies on accurately identifying channel states (line-of-sight vs. non-line-of-sight).
• Estimating channel conditions in indoor wireless local area networks (WLANs) for orthogonal frequency division multiplexing (OFDM) systems typically uses received signal strength identification (RSSI) and channel state information (CSI).

Purpose of the Study:

• To propose and evaluate a novel deep learning architecture for classifying wireless channel states.
• To enhance the accuracy of channel condition estimation in indoor WLANs.

Main Methods:

• A joint convolutional neural network (CNN) and recurrent neural network (RNN) architecture was developed.
• CNNs were employed to extract features from the frequency-domain characteristics of CSI data.
• RNNs were utilized to capture the time-varying features from RSSI and CSI data between packet transmissions.

Main Results:

• The proposed joint CNN-RNN model demonstrated improved classification performance for channel conditions.
• Experimental validation was conducted under realistic indoor propagation environments.
• The novel method achieved a 2% improvement in classification accuracy compared to conventional RNN models.

Conclusions:

• The developed joint CNN-RNN model effectively classifies wireless channel states in indoor environments.
• This approach offers a significant advancement over traditional methods for channel condition estimation.
• The findings contribute to the optimization of location-based services in wireless communication systems.