Evolutionary kernel-based T-spherical fuzzy C-means clustering, application to EMG signal

Clustering complex, real-world datasets characterized by highly overlapping classes, severe noise, and outliers remains a significant challenge in data analysis and pattern recognition. Traditional fuzzy clustering algorithms often fail to capture the deep uncertainty inherent in ambiguous transitional states. To address these limitations, this paper proposes a novel, domain-independent framework: the Evolutionary Kernel T-Spherical Fuzzy C-Means (EKTSFCM). The proposed method integrates T-Spherical Fuzzy Sets (TSFS) to comprehensively model data uncertainty by simultaneously quantifying membership, non-membership, hesitation, and refusal (non-participation) degrees. This makes it particularly effective for isolating outliers and highly ambiguous data points. Furthermore, we introduce a complementary dual-kernel similarity–dissimilarity formulation in which the Gaussian kernel provides a bounded dissimilarity term and the Generalized Bell-shaped kernel provides a similarity term, with the two quantities entering distinct components of the clustering objective to map complex feature spaces and manage boundaries effectively. To mitigate parameter sensitivity and optimize the objective function, an evolutionary optimization algorithm is employed to dynamically tune the dual-kernel and TSFS parameters. Experimental evaluations show that EKTSFCM achieves improved clustering performance relative to the evaluated baseline methods and exhibits greater robustness to severe noise degradation. The proposed framework provides a highly adaptable, mathematically rigorous solution for complex clustering scenarios.