Improved singular value decomposition‐based de‐noising algorithm in digital receiver front‐end

Abstract

It is commonly acknowledged that most of the output signals of digital receiver front-end are noisy and cannot be directly delivered to post-processing subject to the working condition and the limitations of devices. This study presents an improved singular value decomposition-based de-noising algorithm by the adoption of time-domain constrained optimisation. The improved algorithm contains a filter factor which plays a role in distributing information contributions of each retained singular value component to the de-noised signal, according to the relative contribution of noise in each singular value. The selection methods for several critical factors of the algorithm are discussed, including the noisy matrix, de-noising order, noise power and Lagrange multiplier. In particular, the effects of de-noising order and Lagrange multiplier on de-noising performance are studied. The results of the verification tests of extensive simulations and actual measurements from receiver front-end show that the proposed algorithm can significantly reduce the background noise and guarantee the integrity of the information contained in the de-noised signal.