RFI Mitigation in Microwave Radiometry Using Wavelets
The performance of microwave radiometers can be seriously degraded by the presence of radio-frequency interference (RFI). Spurious signals and harmonics from lower frequency bands, spread-spectrum signals overlapping the "protected" band of operation, or out-of-band emissions not properly rejected by the pre-detection filters due to the finite rejection modify the detected power and the estimated antenna temperature from which the geophysical parameters will be retrieved. In recent years,
... recent years, techniques to detect the presence of RFI have been developed. They include time-and/or frequency domain analyses, or statistical analysis of the received signal which, in the absence of RFI, must be a zero-mean Gaussian process. Current mitigation techniques are mostly based on blanking in the time and/or frequency domains where RFI has been detected. However, in some geographical areas, RFI is so persistent in time that is not possible to acquire RFI-free radiometric data. In other applications such as sea surface salinity retrieval, where the sensitivity of the brightness temperature to salinity is weak, small amounts of RFI are also very difficult to detect and mitigate. In this work a wavelet-based technique is proposed to mitigate RFI (cancel RFI as much as possible). The interfering signal is estimated by using the powerful denoising capabilities of the wavelet transform. The estimated RFI signal is then subtracted from the received signal and a "cleaned" noise signal is obtained, from which the power is estimated later. The algorithm performance as a function of the threshold type, and the threshold selection method, the decomposition level, the wavelet type and the interferenceto-noise ratio is presented. Computational requirements are evaluated in terms of quantization levels, number of operations, memory requirements (sequence length). Even OPEN ACCESS Algorithms 2009, 2 1249 though they are high for today's technology, the algorithms presented can be applied to recorded data. The results show that even RFI much larger than the noise signal can be very effectively mitigated, well below the noise level.