Deep Learning-Based LOS and NLOS Identification in Wireless Body Area Networks.

Area of Science:

• Wireless communication
• Machine learning
• Biomedical engineering

Background:

• Ultra-wideband (UWB) technology offers high data rates for Wireless Body Area Networks (WBANs).
• Accurate channel state information is crucial for reliable off-body communication in WBANs.
• Existing methods for classifying line-of-sight (LOS) and non-line-of-sight (NLOS) conditions have limitations.

Purpose of the Study:

• To introduce a novel deep learning (DL) approach for enhanced channel state identification in UWB WBANs.
• To improve the efficiency of distinguishing between LOS and NLOS conditions in off-body communication.
• To validate the proposed DL model using real-world measurement data.

Main Methods:

• Utilizing channel impulse response (CIR) as input for a deep feedforward neural network.
• Developing a DL model trained on CIR data from dynamic indoor WBAN scenarios.
• Comparing the DL approach's performance against existing literature methods.

Main Results:

• The proposed DL approach demonstrates high efficiency in classifying LOS and NLOS conditions.
• Classification accuracy exceeds 98.6% in most tested dynamic indoor scenarios.
• The DL model effectively identifies direct visibility conditions between UWB WBAN nodes.

Conclusions:

• Deep learning provides a highly effective solution for channel condition classification in UWB WBANs.
• The proposed DL method offers superior performance compared to conventional approaches.
• This advancement facilitates more reliable and efficient off-body communication in WBAN applications.