SIGNAL PROCESSING TECHNIQUES FOR TIMEFREQUENCY BASED ANTI-MONEY LAUNDERING SURVEILLANCE

Abstract

Illicit operations may be masked inside lawful financial institutions via money laundering, which is a major problem on a worldwide scale. Complex, dynamic laundering patterns, particularly those containing temporal irregularities and non-stationary transaction behaviours, may be difficult for traditional rulebased anti-money laundering (AML) systems to identify. This research introduces a new method for improving the identification of questionable financial transactions by using time-frequency domain signal processing methods. It is possible to identify and analyse hidden patterns that may indicate money laundering by converting transaction data into time-series signals and then using sophisticated techniques like the ShortTime Fourier Transform (STFT), Wavelet Transform, and Hilbert-Huang Transform. These depictions record the spectral and temporal properties of transaction flows, which allows for the detection of out-of-the-ordinary financial actions. When compared to more traditional models based on statistics and machine learning, the suggested framework performs far better when it comes to spotting anomalies in both simulated and real-world financial information. The merits of timefrequency signal processing as an analytical backbone for next-gen anti-money-laundering monitoring systems are highlighted by this study.