Performance Evaluation of an Automatic GPS Ionospheric Phase Scintillation Detector Using a Machine‐Learning Algorithm

Abstract

Ionospheric phase scintillation can cause errors or outage in GNSS navigation solutions. Timely detection of phase scintillation will enable adaptive processing to mitigate its effects on navigation solutions. This paper presents a machine-learning algorithm to autonomously detect phase scintillation based on frequency domain features. Validation using data from Gakona shows phase scintillation detection accuracy around 92 percent. Test results using data from Poker Flat, Jicamarca, Singapore, and Hong Kong demonstrate the capability of the trained detector to be applied more generally. Performance evaluation reveals that the values of phase scintillation index σ ϕ may be poor indications of scintillation activities. Concurrent phase and amplitude scintillation detection using similar machine-learning algorithms is further investigated with low-latitude data. Results suggest that at low latitudes an amplitude detector alone is sufficient to capture scintillation in general, while at high latitudes, a phase scintillation detector is necessary to capture the dominating phase scintillation events.

Survey on signal processing for GNSS under ionospheric scintillation: Detection, monitoring, and mitigation

“…Some benefits were observed by including a further feature, corresponding to the Fourier Transform of the time series along an observation period of three minutes. A similar approach, targeting phase scintillation, was presented in Jiao, Hall, and Morton (2017b) and Jiao, Hall, and Morton (2017c), leading to detection accuracy around 92%.…”

Automatic detection of ionospheric scintillation‐like GNSS satellite oscillator anomaly using a machine‐learning algorithm

“…SVM is a supervised learning algorithm that deals with classification problems (Bishop, 2006; Cortes & Vapnik, 1995; Jiao et al., 2017). It is known to be one of the best algorithms for solving classification tasks that are not linearly separable.…”

“…Readers are referred to Jiao et al. (2017) for a thorough description and performance analysis of the phase disturbance detection. Note that the phase measurement data is divided into three‐minute windows to ensure there are sufficient samples to capture the phase oscillation.…”

“…It involves a three‐step process. First, phase disturbance events are detected by using the linear support vector machine (SVM)‐based machine‐learning algorithm presented in Jiao, Hall, and Morton (2017). The triple‐frequency carrier‐phase measurement segments are identified based on the linear SVM detection outputs.…”

Semi-Supervised GNSS Scintillations Detection Based on DeepInfomax

“…Such approach was originally proposed in [8,9], where the authors use an automatic scintillation detector based on a supervised machine learning technique called support vector machine (SVM), respectively for amplitude and phase scintillation. This approach provides good detection results but introduces a Fast Fourier Tranform (FFT) operation on the data over windows of 3 min, thus reducing the temporal resolution of the detection results.…”