A Novel Ensemble Empirical Decomposition and Time–Frequency Analysis Approach for Vibroarthrographic Signal Processing

Abstract

Signal processing techniques play a critical role in addressing real-world applications across domains such as sensor analysis, defence, and clinical and biomedical fields. Within healthcare, computer-aided diagnostic (CAD) systems have become pivotal in supporting medical professionals with the interpretation of data and images, especially in medical imaging and radiological diagnostics. For diagnosing joint disorders, both time-domain and frequency-domain analyses are employed to examine complex, non-stationary, and nonlinear signals. To process Vibroarthrographic signals in this context, an initial step involves applying the Hilbert-Huang Transform, which comprises two stages: Empirical Mode Decomposition (EMD) for computing intrinsic mode functions (IMFs), followed by the Hilbert transform for further signal analysis. In our proposed approach, we utilized Complete Ensemble Empirical Mode Decomposition with Adaptive Noise and Time-Varying Frequency Empirical Mode Decomposition (TVF-EMD) to compute IMFs, as well as Variation Mode Decomposition to calculate mode signals. Subsequent feature extraction incorporates both time and frequency characteristics, focusing on metrics such as pixel intensity, mean, and standard deviation. These features then serve as inputs to machine learning models for classification tasks, distinguishing between healthy and non-healthy signal samples. In our model, we employed a Least Squares Support Vector Machine (LS-SVM) and a Support Vector Machine with Recursive Feature Elimination (SVM-RFE) to enhance classification accuracy. This sequence of signal processing and machine learning steps demonstrates a structured and effective approach for CAD-based diagnosis in joint disorder assessments.