Concepedia

Interference Source Identification for IEEE 802.15.4 wireless Sensor Networks Using Deep Learning

Yi Su, Hao Wang, Wenqian Xue, Xiaojing Fan, Lefei Wang, Ryuichi Matsukura

2018 · 18 citations · 12 references

TL;DR

Interference in the unlicensed band degrades performance and connectivity of sensor nodes. The study proposes a real‑time deep‑learning method to classify external interference sources in an 802.15.4 wireless sensor network. The approach trains a convolutional neural network on RSSI samples collected in an office setting, labels data with wireless sniffers, and employs micro‑ and macro‑level models to predict interference types such as Wi‑Fi, WLAN, BLE iBeacon, and microwave oven. The trained network, implemented on both an IEEE 802.15.4 SoC and a Linux system, achieves high‑accuracy classification of major interference sources.

Abstract

Due to the interference issue in unlicensed band, sensor nodes frequently encounter degraded performance or lack of connection. This paper provides a real-time external interference source classification method for an 802.15.4-based wireless sensor network using deep learning. It uses RSSI sampling for collecting training data as well as online test data in an office environment. The output interference source type includes Wi-Fi beacon, different classes of WLAN traffic, BLE iBeacon, and microwave oven. Wireless sniffers are used to help labeling the ground truth of the sample data. We have trained a deep neural network with two hidden convolutional layers using raw RSSI samples as inputs. A micro-level model and a macro-level model are provided to predict the interference source based on the deep learning result. With implementation on both IEEE 802.15.4 SoC and Linux-based system, our experimental results show that the proposed framework can classify the major interference types with high accuracy.