This paper introduces a new segmentation algorithm of voltage waveforms. This new algorithm uses the strength of the tensor concept of the voltage signals to generate a unique pattern of deformation in three-phase systems. This deformation pattern is used to perform segmentation by detecting residues. In the process of segmentation, a modification to the Kalman filter has been proposed based on the work presented by Rakhee Panigrahi and Cesar Duarte. With this modification, we have obtained a robust response with a term included for estimating adaptively to sudden changes in the system. Additionally, it has been implemented a strategy of adaptive thresholding to adapt to noisy signals. This segmentation strategy is validated using a set of synthetic signals with variation of the remnant voltage and the starting point of the disturbance. Finally, the proposed algorithm is tested with a set of real signals.