Analysis of Microwave Reflectometry Data Using Empirical Mode Decomposition for Beat Frequency Estimation

In Fusion plasma diagnostics based on microwave reflectometry, the tokamak environment is highly noisy and turbulent. This makes it difficult to measure the beat frequency accurately. Traditional methods reported to remove noise involve filtering, wavelet decomposition etc. Since the Phase information is important for getting an accurate density profile, the noise removal methods employed should not alter the signal's phase. In this paper, Empirical Mode Decomposition (EMD) followed by energy-based thresholding and correlation-based thresholding are proposed for selecting the meaningful Intrinsic Mode Functions (IMFs). The results obtained from these two methods are compared based on the resultant signal's phase and certain evaluation metrics like mean squared error (MSE), mean absolute error(MAE) and peak signal-noise ratio (PSNR(dB)). The analysis is carried out with the ADITYA -Upgrade tokamak back wall as a target, and it can be further extended to the plasma scenario. The values of MSE, MAE and PSNR are 0.0149, 0.1003, 7.8474 and 0.0139, 0.0098, and 7.8481, respectively, for correlation-based and Energy-based thresholding methods of IMF selection respectively.